A Configuration Theory Assessment of Marketing Organization Fit with Business Strategy and Its Relationship with Marketing Performance 

Theory posits that organizing marketing activities in ways that fit the implementation requirements of a business's strategy enhances performance. However, conceptual and methodological problems make it difficult to empirically assess this proposition in the holistic way that it is theoretically framed. Drawing on configuration theory approaches in management, the authors address these problems by assessing marketing organization fit with business strategy as the degree to which a business's marketing organization differs from that of an empirically derived ideal profile that achieves superior performance by arranging marketing activities in a way that enables the implementation of a given strategy type. The authors suggest that marketing organization fit with strategic type is associated with marketing effectiveness in prospector, defender, and analyzer strategic types and with marketing efficiency in prospector and defender strategic types. The study demonstrates the utility of profile deviation approaches for strategic marketing theory development and testing.

Most businesses find it easier to formulate strategies that outline how they intend to achieve their goals than how to implement them (e.g., Noble and Mokwa 1999; Walker and Ruekert 1987). The literature suggests that an important cause of such strategy implementation difficulties is the way businesses organize their marketing activities (e.g., McKee, Varadarajan, and Pride 1989; Ruekert and Walker 1987). In particular, marketing theory posits that to enable strategy implementation and achieve superior performance, managers should organize marketing activities in different ways depending on their business strategy (e.g., Slater and Olson 2000; Walker and Ruekert 1987). However, organizing marketing activities in ways that successfully enable business strategy implementation is recognized as one of the most difficult challenges facing managers (e.g., Cespedes 1995; Webster 1997). Yet despite the theoretical and managerial importance of this issue, researchers know little about how marketing activities should be organized to enable business strategy implementation or how this affects performance (e.g., Walker and Ruekert 1987; Workman, Homburg, and Gruner 1998).

Investigating this complex theoretical and managerial problem presents two significant challenges. First, the organization of marketing activities and business strategy are each viewed as multidimensional phenomena consisting of many different but interconnected elements (e.g., Walker and Ruekert 1987). Yet strategic marketing theory frames relationships between these phenomena in holistic terms as marketing organization's role in implementing business strategy. Therefore, evaluating this relationship in these holistic terms requires a simultaneous assessment of the relationships between the many variables making up marketing organization and business strategy (e.g., Slater 1995; Walker and Ruekert 1987). Second, successfully organizing marketing activities to implement business strategy involves reconciling multiple and often conflicting contingencies (e.g., Ruekert, Walker, and Roering 1985). The wide range of possible contingencies makes the identification of "correct" configurations of marketing organization variables needed to implement a particular business strategy extremely difficult. Therefore, the challenge facing researchers is how to assess whether marketing activities are organized in ways that enable the implementation of a particular business strategy (e.g., Walker and Ruekert 1987).

Although assessing many strategic marketing theory propositions involves similar challenges, available research approaches in marketing are not well suited to deal with these problems. Surprisingly, the marketing literature does not address how such problems can be solved. Fortunately, research developments in organization theory (e.g., Powell 1992; Venkatraman and Prescott 1990) and strategic management (e.g., Doty, Glick, and Huber 1993; Ketchen et al. 1997) provide approaches appropriate for assessing such complex theoretical relationships. In this article, we draw on these developments to build and empirically assess a conceptual model that links the degree to which marketing activities are organized in ways that enable business strategy implementation with two different marketing performance outcomes.

Within this important domain, our study makes two contributions. First, we fill a major knowledge gap by providing empirical support for theorized links among the organization of marketing activities, business strategy, and marketing performance. This helps managers understand how to organize marketing activities to meet the implementation requirements of different business strategies and why this is important in driving performance. Second, we introduce to the marketing literature a method for testing relationships involving the simultaneous assessment of multiple interrelated variables. We demonstrate that this method provides researchers with a way to empirically assess relationships involving complex, multidimensional phenomena that is more consistent with the holistic framing of strategic marketing theory than traditional approaches are (e.g., Meyer, Tsui, and Hinings 1993).

Assessing whether a business's marketing activities are organized in ways that enable the implementation of its strategy and the impact this has on performance requires the simultaneous consideration of multiple characteristics of the business (e.g., Doty, Glick, and Huber 1993). In addressing similar research questions, scholars in organization theory and strategic management have used configuration theory- based approaches (e.g., Miller 1997; Veliyath and Srinivasan 1995). A configuration denotes a multidimensional constellation of the strategic and organizational characteristics of a business (e.g., Meyer, Tsui, and Hinings 1993; Miller and Mintzberg 1988). Configuration theory posits that for each set of strategic characteristics, there exists an ideal set of organizational characteristics that yields superior performance (e.g., Van de Ven and Drazin 1985). These configurations are ideal because they represent complex "gestalts" of multiple, interdependent, and mutually reinforcing organizational characteristics that enable businesses to achieve their strategic goals (e.g., Ketchen, Thomas, and Snow 1993; Miller 1997). Our research question pertains to the relationship between the marketing organization configuration and the business's strategy, rather than the coalignment of variables within the marketing organization configuration. Therefore, in Figure 1, we combine insights from configuration theory and the marketing literature to develop a conceptual model that links the degree to which marketing activities are organized in ways that enable business strategy implementation with marketing performance.

Defining and Assessing Marketing Organization Fit with Business Strategy

Marketing scholars have used many different terms--including "match," "alignment," "congruence," "complementary," and "consistency"--to denote holistic relationships between multidimensional phenomena such as marketing organization and business strategy. Although each of these terms can connote different meanings and technical specifications, they are often used interchangeably. To more precisely specify and assess such relationships, configuration theory-based studies draw on the well-developed literature regarding fit. In this literature, fit is recognized as a term that can be defined in several ways, each of which has specific implications for how relationships between variables are conceptualized and tested (Powell 1992; Venkatraman and Camillus 1984). Therefore, management scholars have specified the different conceptualizations and technical specifications of alternative definitions of fit and have developed guidelines for selecting the approaches that are most appropriate in studying different kinds of research questions (e.g., Venkatraman 1989). This literature specifies that when fit among multiple variables is considered simultaneously (as in the holistic study of the relationship between organization and strategy) and the impact on criterion variables (e.g., performance) is assessed, fit should be conceptualized and assessed as "profile deviation" (e.g., Doty, Glick, and Huber 1993; Venkatraman 1990).

A profile deviation approach views fit between organization and strategy in terms of the degree to which the organizational characteristics of a business differ from those of a specified profile identified as ideal for implementing a particular strategy (e.g., Venkatraman 1989; Zajac, Kraatz, and Bresser 2000). Ideal profiles are defined as configurations of organizational characteristics that fit with the implementation requirements of a particular strategy and thus produce high performance (e.g., Gresov 1989; Venkatraman and Prescott 1990). From this perspective, marketing organization fit with business strategy can be defined as the degree to which a business's marketing organization profile differs from that of an ideal marketing organization that achieves superior performance by arranging marketing activities in a way that enables the implementation of a given business strategy.

Ideal profiles against which fit can be assessed may be determined either theoretically or empirically (e.g., Venkatraman 1990; Zajac, Kraatz, and Bresser 2000). Developing theoretically derived ideal profiles requires that the relevant theoretical literature be sufficiently detailed to enable precise numerical scores to be estimated for the appropriate set of dimensions in the ideal profile (e.g., Drazin and Van de Ven 1985). However, as the configuration literature acknowledges, there are few domains in which existing theoretical knowledge is sufficiently detailed to enable researchers to objectively translate theoretical statements from the literature into precise numerical estimates across multiple dimensions (e.g., Gresov 1989; Venkatraman 1989). In the marketing domain, existing theory indicates some marketing organization characteristics that may be appropriate for firms pursuing certain types of strategy (e.g., Ruekert and Walker 1987). However, from an ideal profile perspective, the specifications provided by marketing theory are not sufficiently detailed to enable estimation of numerical scores, nor do they consider many of the organization characteristics and types of strategy identified as important in assessing marketing organization fit with business strategy.

In this common circumstance, when ideal profiles cannot be precisely specified from existing theory, the configuration literature advocates assessing fit with empirically derived ideal profiles (e.g., Gresov 1989; Ketchen, Thomas, and Snow 1993). In the context of marketing organization fit with business strategy, this approach requires the identification of high-performing businesses implementing a given strategy and a calibration of their marketing organization characteristics as an ideal profile for implementing that strategy (e.g., Drazin and Van de Ven 1985; Venkatraman and Prescott 1990). These businesses are considered to have ideal profiles because their superior performance indicates that they have configured their marketing organization in a way that enables superior implementation of their business strategy (e.g., Van de Ven and Drazin 1985).

Configurational Elements of Marketing Organization Fit with Business Strategy

As illustrated in Figure 1, configuration theory and the marketing literature suggest two major constructs that are relevant to understanding and assessing marketing organization fit with business strategy: business's strategic type and marketing's organizational characteristics. Strategic type pertains to the planned patterns of organizational adaptation to the market through which a business seeks to achieve its strategic goals (e.g., Conant, Mokwa, and Varadarajan 1990; Matsuno and Mentzer 2000). Miles and Snow (1978, p. 29) identify three viable strategic types, which differ primarily in terms of product-market strategy choices (e.g., Slater and Narver 1993; Walker and Ruekert 1987).[ 1] Prospector strategic types proactively seek and exploit new market opportunities and often experiment with responses to changing market trends. They aggressively compete on innovation, seeking first-mover advantages from developing new offerings and pioneering new markets. Defender strategic types focus more narrowly on maintaining a secure position in existing product-markets. They often compete through operations or quality-based investments that offer efficiency related advantages, rarely pioneering the development of new markets or products. Analyzer strategic types balance a focus on securing their position in existing core markets with incremental moves into new product markets. They compete by balancing investments in creating differentiation-based advantages with operating efficiency.

Marketing's organizational characteristics are the many important structural and task characteristics that together constitute the way marketing activities are organized within the business (Day 1997; Workman, Homburg, and Gruner 1998). The structural characteristics of the marketing organization pertain to how marketing activities and related decision-making authority are arranged (e.g., Doty, Glick, and Huber 1993; Ruekert, Walker, and Roering 1985). Although the literature identifies several different structural characteristics of organization, three have been viewed as particularly important in previous marketing strategy research: centralization regarding the concentration of decision-making authority at higher levels of the business's hierarchy (e.g., Jaworski and Kohli 1993; Moorman, Deshpande, and Zaltman 1993); formalization, which is the degree to which standardized rules and procedures proscribe how marketing activities are performed (Olson, Walker, and Ruekert 1995; Workman, Homburg, and Gruner 1998); and specialization, which is the extent to which marketing activities are narrowly divided into unique elements that are performed by those with specialized knowledge (e.g., Walker and Ruekert 1987). Together, these structural characteristics indicate whether marketing activities are arranged in a bureaucratic or an organic manner (Moorman and Miner 1997; Ruekert, Walker, and Roering 1985).

The task characteristics of marketing organization pertain to the nature of the marketing activities undertaken and the ways they are performed (e.g., Day 1999; Ostroff and Schmitt 1993). Among the different task characteristics identified in the literature, three have been viewed as important both in configuration theory studies in management and in previous marketing strategy studies: task complexity, which is the extent of variability in marketing activities undertaken and the degree to which they can be easily performed (Menon and Varadarajan 1992; Olson, Walker, and Ruekert 1995); marketing capabilities regarding the business's ability to perform common marketing work routines through which available resources are transformed into valuable outputs (e.g., Bharadwaj, Varadarajan, and Fahy 1993; Day 1994; Webster 1997); and work group interdependence, which is the degree to which workflows within the business require cooperation between teams in performing marketing activities (e.g., Ruekert and Walker 1987; Van de Ven, Delbecq, and Koenig 1976). Together, these task characteristics indicate the ability of the marketing organization to perform necessary marketing activities and the degree to which team-based workflows are needed to accomplish them.

Marketing Organization Fit with Strategic Type and Performance

Fit between the organizational characteristics of a business and its strategic type is viewed as a desirable state that leads to superior performance (e.g., Miles and Snow 1994; Porter 1996). Marketing theory suggests that this is also true of fit between marketing organization characteristics and strategic type. For example, the literature indicates that the marketing activities needed to implement each strategic type are different and that successfully accomplishing these marketing activities requires marketing organizations with different configurations of structural and task characteristics (e.g., Matsuno and Mentzer 2000; McKee, Varadarajan, and Pride 1989; Walker and Ruekert 1987). Therefore, marketing theory suggests that organizing marketing activities in ways that fit the business's strategic type is an important driver of marketing performance outcomes (e.g., Walker and Ruekert 1987).

Furthermore, resource-based view theory indicates that fit between marketing organization characteristics and strategic type may also exhibit the inimitability and nonsubstitutability characteristics identified as essential for sustaining competitive advantage. For example, if a firm's superior performance is driven by marketing organization fit with strategic type, it will be difficult for competitors to identify the source of the firm's performance superiority (e.g., Barney 1991). Even if identified as a driver of superior performance, the ability of competitors to distinguish precisely how this is accomplished is limited, making imitation difficult (e.g., Bharadwaj, Varadarajan, and Fahy 1993; Day 1994). In addition to being difficult to imitate, the marketing literature suggests that there may be no substitute for marketing organization fit with strategic type in driving marketing performance (e.g., Moorman and Rust 1999; Workman, Homburg, and Gruner 1998). Therefore, marketing and resource-based view theory suggest that marketing organization fit with strategic type leads to superior marketing performance and that this can be sustained over time (e.g., Powell 1992; Walker and Ruekert 1987).

However, what constitutes superior marketing performance may be different in different businesses. For example, organization theory posits that effectiveness, regarding the degree to which desired organizational goals are achieved, and efficiency, regarding the ratio of organizational resource inputs consumed to goal outcomes achieved, are two important and distinct dimensions of organizational performance (e.g., Bonoma and Clark 1988; Lewin and Minton 1986). The literature suggests that because these two dimensions of performance may not converge and can even be inversely related in the short run (e.g., Bhargava, Dubelaar, and Ramaswami 1994), firms make important trade-off decisions in emphasizing either effectiveness or efficiency in their marketing goal setting and resource allocations (e.g., Morgan, Clark, and Gooner 2002; Walker and Ruekert 1987). Therefore, configuration theory suggests that the ideal marketing organization required to fit with a particular strategic type differs depending on whether the firm seeks to maximize either effectiveness or efficiency (e.g., Tsui 1990). Assessing marketing organization fit with strategic type requires an identification of different ideal profiles against which to assess fit for firms seeking to maximize either the effectiveness or the efficiency dimension of their marketing performance (e.g., Ostroff and Schmitt 1993; Walker and Ruekert 1987).

In developing hypotheses of expected relationships between marketing organization fit with strategic type and its performance outcomes, we draw directly on existing theory and empirical evidence when possible. However, although many studies have investigated structural characteristics of marketing organization (e.g., Ruekert, Walker, and Roering 1985; Workman, Homburg, and Gruner 1998), few studies have investigated marketing organization task characteristics. Similarly, existing knowledge of analyzer strategic types is less developed in the marketing literature than knowledge regarding prospector and defender types. In addition, with a few notable exceptions (e.g., Walker and Ruekert 1987), strategic marketing theory has not considered how seeking to maximize different dimensions of marketing performance affects marketing organization. Therefore, we draw on a necessarily broad reading of the literature in developing our hypotheses.

Marketing Organization Fit with Strategic Type and Marketing Effectiveness

Marketing effectiveness pertains to the degree to which desired market-based goals are achieved (e.g., Clark 2000; Morgan, Clark, and Gooner 2002). Theory suggests that for effectiveness-maximizing businesses of each strategic type, an ideal marketing organization exists in which the configuration of structural and task characteristics enables the implementation of the business's strategy in a way that leads to superior marketing effectiveness (e.g., Cespedes 1991; Day 1997; Ruekert and Walker 1987). For example, defender strategic types focus on maintaining secure positions in established markets. Therefore, implementing this strategy requires a marketing organization configured to achieve needed market-based goals through performance of routine tactical marketing activities (e.g., Ruekert and Walker 1987; Slater and Narver 1993). Performing such routine activities calls for a marketing organization with a highly centralized, formalized, and unspecialized structure and a relatively narrow range of marketing capabilities (e.g., Walker and Ruekert 1987). By narrowly focusing the deployment of available resources, marketing organizations with these characteristics may benefit from greater depth in a few key marketing capabilities. This may be leveraged through centralized authority structures that provide control over the focus of future resource deployment and formalized work routines that minimize errors in executing required activities. Organizing marketing activities in this way should enable a business implementing a defender strategy to achieve superior marketing effectiveness.

Conversely, prospector strategic types focus on entering unfamiliar new markets and attaining differentiation-based advantages. Therefore, achieving required marketing goals in implementing a prospector strategy involves performing many complex marketing activities (e.g., McDaniel and Kolari 1987; McKee, Varadarajan, and Pride 1989). Accomplishing these activities ideally requires specialized, decentralized, and informal marketing structures with team work-flows and a range of strong marketing capabilities (e.g., Ruekert, Walker, and Roering 1985; Walker 1997). In implementing prospector strategies, such organizational characteristics should enhance marketing effectiveness because they empower marketing specialists with access to wide-ranging capabilities and provide the decision-making freedom and work routine flexibility to use these capabilities to provide timely and innovative responses in dynamic product-markets (e.g., Walker and Ruekert 1987).

Businesses pursuing analyzer strategies operate in a range of established and new markets and seek to attain both cost and differentiation-based advantages. Therefore, analyzer strategic types require marketing organizations that are able to achieve needed marketing goals by performing a particularly wide and dynamic range of marketing activities (e.g., Slater and Narver 1993). Marketing organizations ideal for the analyzer strategic type in effectiveness maximizing businesses should therefore have high levels of marketing specialization, but formalized and centralized structures with strong marketing capabilities and team workflows (e.g., Miles and Snow 1994). Such specialization, team workflow, and marketing capability characteristics enable businesses implementing an analyzer strategy to respond quickly to the complex marketing activity requirements of unfamiliar markets while continuing to service the more routine demands of established markets. At the same time, formalization minimizes error in performing required marketing activities, and centralization allows tight control over the new market opportunities pursued. Marketing organizations with such ideal characteristics should enable the implementation of analyzer strategies in a way that produces superior marketing effectiveness.

In summary, we expect a business's marketing effectiveness to be greater when its marketing organization characteristics are similar to those of the effectiveness-maximizing ideal profile in which marketing activities are arranged to fit the implementation requirements of the business's strategic type in ways that enable marketing goals to be achieved. Therefore, we hypothesize that

H₁: The more similar a business's marketing organization profile is to that of the ideal marketing organization for its strategic type, the greater is its marketing effectiveness.

Marketing Organization Fit with Strategic Type and Marketing Efficiency

Marketing efficiency is the ratio of marketing performance outcomes achieved to resource inputs consumed (e.g., Bonoma and Clark 1988; Morgan, Clark, and Gooner 2002). Theory suggests that for efficiency-maximizing businesses of each strategic type, there exists an ideal marketing organization in which the configuration of structural and task characteristics enables the implementation of the business's strategy in a way that leads to superior marketing efficiency (e.g., Jennings and Seaman 1994; Milgrom and Roberts 1995; Ruekert and Walker 1987). For example, implementing defender strategies requires achieving cost-based advantages in established markets. Creating specialized structures with team workflows and developing a wide range of strong marketing capabilities are not likely to be efficient ways to achieve marketing goals when implementing this strategy (e.g., Conant, Mokwa, and Varadarajan 1990). Available resources are more productively deployed in simplifying marketing activities, increasing structural formalization and centralization, and developing a narrow range of marketing capabilities (e.g., Slater and Narver 1993). Such ideal marketing organization characteristics should maximize marketing efficiency in implementing a defender strategy by allowing more focused resource deployment in capability building, greater control of decisions involving future resource allocation, and the efficiency benefits of increased routinization (e.g., Walker and Ruekert 1987).

Conversely, efficiently implementing prospector strategies requires achieving marketing goals by entering unfamiliar new markets and delivering differentiation-based competitive advantages while minimizing the resources consumed (e.g., McKee, Varadarajan, and Pride 1989). One way to accomplish this is to narrow the scope of product-market opportunities explored and competitive advantage pursued. This simplifies required marketing activities and enables more focused investment in a narrower range of marketing capabilities. Needed flexibility to respond quickly to appropriate new opportunities can be provided by empowering marketing personnel with decentralized and informal structures. By maintaining response flexibility while consuming fewer resources through more focused investments in marketing capabilities, ideal marketing organizations that fit the prospector strategic type in this way should be more efficient in achieving marketing goals (e.g., Miles and Snow 1994).

Efficiently implementing an analyzer strategy requires minimizing the resources consumed to accomplish needed marketing goals in the different types of product-markets in which analyzers operate. Here, an ideal marketing organization requires sufficient levels of specialization, team work-flows, and marketing capabilities to perform the complex range of different marketing activities required to achieve marketing goals (e.g., Slater and Narver 1993). However, to minimize the resources consumed, such analyzer businesses may also choose to operate in product-markets and pursue competitive advantages that only require strength in a narrower range of marketing capabilities. At the same time, accomplishing needed marketing goals efficiently also requires marketing organizations with enough centralized authority to tightly control resource allocations and sufficient formalization to benefit from routinization wherever possible (see Miles and Snow 1994). Marketing organizations with such ideal characteristics in businesses that implement analyzer strategies should be more efficient.

To summarize, we expect that a business's marketing efficiency will be greater when its marketing organization characteristics are similar to those of the efficiencymaximizing ideal profile in which marketing activities are arranged to fit the implementation requirements of the business's strategic type in ways that minimize the resources consumed. We therefore hypothesize that

H₂: The more similar a business's marketing organization profile is to that of the ideal marketing organization for its strategic type, the greater is its marketing efficiency.

In examining fit-performance relationships, the configuration theory literature advocates the use of single industry studies to control for industry effects and isolate more effectively the relationships of interest (e.g., Dess, Newport, and Rasheed 1993; Ketchen, et al. 1997). We selected the trucking industry as appropriate for studying marketing organization fit with strategic type and its relationship with marketing performance for several reasons. First, with more than $372 billion spent annually, accounting for some 6% of gross domestic product, and 9.5 million people directly employed, trucking is a large and important industry in the United States (e.g., American Trucking Association 1999). Second, since deregulation in 1980, trucking has become a dynamic and competitive industry (e.g., Silverman, Nickerson, and Freeman 1997) in which effective and efficient marketing has become an important driver of firm performance (e.g., Lambert, Lewis, and Stock 1993; MacLeod et al. 1999). Third, the industry contains many single business firms, which reduces the problems associated with relating business unit-level phenomena and corporate-level performance data (e.g., Ketchen et al. 1997). Fourth, the trucking industry is relatively fragmented, providing a large population of firms for sampling purposes (e.g., Boyer 1993). Fifth, because of federal reporting requirements, objective performance data for the trucking industry are available, which reduces the dangers of common method bias associated with collecting data on independent and dependent variables from the same source (e.g., Olson, Walker, and Ruekert 1995; Venkatraman and Ramanujam 1986).

We collected primary data using a key-informant survey design. We mailed questionnaires to the chief marketing executive (CME) of 677 businesses, which we randomly selected from the 2771 listed in the Transportation Technical Services (TTS) database. The TTS database is representative of the industry, listing businesses generating more than 97% of total intercity freight revenues (U.S. Bureau of the Census 1998). Of the 660 deliverable surveys, 217 were completed and returned. Of the returned surveys, 8 were unusable, resulting in an effective response rate of 31%.[ 2] To ensure comparability, we deleted observations from the data set in which complete sets of both primary questionnaire data and secondary performance data were not available. The final data set contained 186 businesses, of which 28% reported sales of less than $10 million, 25% reported sales of $10-$25 million, 23% reported sales of $26-$65 million, and 24% reported sales greater than $65 million. Of these businesses, 77 were pursuing a defender strategy, 45 were pursuing a prospector strategy, and 64 were pursuing an analyzer strategy.

Measures

Several different operationalization alternatives exist for some of the constructs in our study. In these situations, we selected the operationalization with the strongest measurement history in the literature and the greatest face validity with managers in our pretest. The items used in our scales appear in Appendix A, and we discuss them subsequently, with descriptive statistics presented in Table 1.

Strategic type. We operationalized Miles and Snow's (1978) strategic types using the self-typing paragraph descriptor approach (e.g., Doty, Glick, and Huber 1993). This measure focuses exclusively on business strategy and excludes the structural and system elements elaborated in Miles and Snow's broader descriptions of ideal organizational archetypes. This measure has been widely used as an indicator of strategic type by marketing and management researchers (e.g., Matsuno and Mentzer 2000; McDaniel and Kolari 1987; Zahra and Pearce 1990) and has been demonstrated to yield valid measures (e.g., James and Hatten 1995; Shortell and Zajac 1990).

Marketing's organizational characteristics. We measured the structural characteristics, centralization and formalization, with multi-item scales adapted from Deshpande and Zaltman (1982), Jaworski and Kohli (1993), and John and Martin (1984), based on the well-known organization theory scales developed by Aiken and Hage (1968). We operationalized specialization using a scale adapted from Doty, Glick, and Huber (1993), based on an organization theory scale developed by Inkson, Pugh, and Hickson (1970).

Task characteristics. We measured task complexity using a scale adapted from Doty, Glick, and Huber (1993), and we assessed work group interdependence using Van de Ven, Delbecq, and Koenig's (1976) measure. We developed new measures of marketing capability for this study, combining insights from the literature with interviews with trucking industry experts and senior marketing personnel. We identified and assessed two types of marketing capabilities: specialized capabilities regarding the specific marketing mix-based work routines used to transform available resources into valuable outputs (e.g., Day 1994; Grant 1996) and architectural capabilities regarding the marketing strategy formulation and execution work routines used to develop and coordinate specialized capabilities and their resource inputs (e.g., Bharadwaj, Varadarajan, and Fahy 1993; Day 1997). We measured these marketing capabilities with scales that assessed how well businesses performed five specialized marketing activities and four architectural marketing activities compared with their competitors (e.g., Conant, Mokwa, and Varadarajan 1990).

Marketing performance. We assessed marketing effectiveness using a perceptual measure with items tapping the degree to which the firm achieved its market share growth, sales growth, and market position goals (e.g., Clark 2000). We calculated marketing efficiency as the ratio of marketing and selling expenses to the firm's gross operating revenue using objective secondary financial data from TTS (e.g., Bonoma and Clark 1988).

Psychometric Analyses

We assessed the measurement properties of the constructs using confirmatory factor analyses (CFAs). Because of the relatively small number of observations, we divided the measures into three subsets of theoretically related variables in line with our conceptual model (e.g., Kohli and Jaworski 1994; Moorman and Miner 1997). This ensured that our CFAs did not exceed the five-to-one ratio of parameter estimates to observations recommended in the literature (Bentler and Chou 1987). The three measurement models fit well as indicated by the CFA results for the three structural characteristics constructs (&chi² = 59.42, degrees of freedom [d.f.] = 41, p = .03, goodness-of-fit index [GFI] = .95, root mean square error of approximation [RMSEA] = .05), the four task characteristics (&chi ² = 111.80, d.f. = 85, p = .03, GFI = .93, RMSEA = .04), and the two marketing performance constructs (χ² = .32, d.f. = 2, p = .85, GFI = .99, RMSEA = .01). When significant correlations were observed between constructs (Table 1), we also conducted additional pairwise discriminant validity assessments. This involved comparing χ ² statistics in measurement models in which the covariance coefficient between the two constructs was allowed to vary and then fixed at one (Anderson and Gerbing 1988; Bagozzi and Phillips 1982). Changes in χ²were large in each of the pairwise tests, suggesting discriminant validity in each model. Reliability analyses for the measures produced Cronbach's alpha values ranging from .72 to .85 (Table 1), suggesting acceptable reliability for all constructs.

Analysis of nonresponse bias by means of an extrapolation approach (Armstrong and Overton 1977) revealed no significant differences between first wave (early) and second wave (late) respondents on any of the constructs. This suggests that nonresponse bias is unlikely to be present in the data. To validate the data provided by the key marketing respondents, we sought additional data from a second respondent (see Slater 1995). For each of the 186 firms responding to the CME survey, we also sent the chief executive officer, president, or other general manager (GM) level executive a questionnaire containing replicated scales on the marketing organization and performance measures. A total of 88 of the GM level executives responded, producing an effective response rate of 47%. For each construct, we assessed the validity of the key informant data by examining mean scores, correlations, and paired t-tests for the GM and CME level responses (e.g., Hughes and Garrett 1990). As shown in Table 2, the significant inter-rater correlations and insignificant mean differences with no systematic bias in direction between raters support the validity of the key informant data (see Jaworski and Kohli 1993).

Testing Configuration Theory Predictions with Profile Deviation Analysis

Testing the hypothesized relationships involved several stages. First, we standardized the data to remove the effects of different measurement units (e.g., Gresov 1989). Second, we identified ideal marketing organization profiles against which marketing organization fit with strategic type could be assessed (e.g., Doty, Glick, and Huber 1993). Consistent with established configuration theory procedures, we identified the highest performing businesses of each strategic type on each of the marketing performance variables and calibrated the marketing organization characteristics of these high performers as the ideal marketing organization profiles (e.g., Doty, Glick, and Huber 1993; Drazin and Van de Ven 1985; Venkatraman 1990).

Profile deviation studies typically select the highest performing 10% or 15% of businesses in a data set to calibrate ideal profiles (e.g., Van de Ven and Drazin 1985, Venkatraman and Prescott 1990). To select the appropriate number of top performers for our study, we examined scree plots of the marketing effectiveness (in testing H1) and marketing efficiency (in testing H2) performance of the businesses in our data set. These indicated a drop-off in performance after the top five marketing effectiveness performers and the top five marketing efficiency performers for each of the three strategic types. Therefore, we selected the five highest marketing effectiveness performers of each strategic type to calibrate the ideal marketing organization profiles for effectivenessmaximizing businesses and the five highest marketing efficiency performers of each strategic type to calibrate the ideal marketing organization profiles for efficiencymaximizing businesses. Consistent with marketing (e.g., Ruekert and Walker 1987) and configuration theory predictions (e.g., Tsui 1990), we found that only one firm was identified as both a top effectiveness and efficiency performer in our sample.

In testing H₁,we calculated the mean scores of the top marketing effectiveness performers for each strategic type on each of the seven marketing organization variables to form the ideal marketing organization profiles (e.g., Venkatraman 1989). For the remaining firms, we calculated the Euclidean distance of each firm from the ideal marketing organization profile for its strategic type across the seven dimensions, representing the seven marketing organization variables (e.g., Drazin and Van de Ven 1985; Venkatraman 1990), as follows:

Multiple line equation(s) cannot be represented in ASCII text; please see PDF of this article if available. where

X_sj = the score for a firm in the study sample on the jth dimension,

[Multiple line equation(s) cannot be represented in ASCII text; please see PDF of this article if available.] X_ij = the mean for the ideal profile along the jth dimension, and

j = the number of profile dimensions ( 1, 2, ..., 7).

This calculation provides a profile deviation score that represents the degree to which the marketing organization profile of each firm is similar to that of the ideal profile for its strategic type. The profile deviation score for each firm was then regressed onto marketing effectiveness to test H1. Because the trucking industry is a capital-intensive service business, in which economies of scale may be expected to affect performance (e.g., Bharadwaj, Varadarajan, and Fahy 1993; Boyer 1993; Silverman, Nickerson, and Freeman 1997), we also included firm size, indicated by the natural logarithm of the number of employees, in our regression equations as a control variable (e.g., Germain, Droge, and Daugherty 1994). We then repeated this procedure using the top marketing efficiency performers for each strategic type in calibrating ideal profiles, with the profile deviation scores of the remaining firms regressed onto marketing efficiency to test H₂.

For our hypotheses to be supported, empirical results should indicate that deviation from the ideal marketing organization profile is negatively and significantly related to marketing effectiveness and positively and significantly related to marketing efficiency for each of the three strategic types (e.g., Drazin and Van de Ven 1985; Gresov 1989).[ 3] Assessing the power of these hypothesis tests requires comparing the regression models containing deviation from the ideal marketing organization profile with regression models containing deviation from an alternative "nonideal" baseline profile (e.g., Venkatraman 1989). Therefore, we randomly selected five firms of each strategic type for each of the performance dimensions, in which the level of marketing organization fit with strategic type was unknown. We used these randomly selected firms to calibrate an alternate nonideal profile from which we calculated the deviation of the remaining firms (e.g., Venkatraman and Prescott 1990). We then substituted the nonideal profile deviations into the regression models in place of the ideal profile deviations to enable comparisons.

Before testing the hypotheses, we first validated two assumptions implicit in our conceptualization. First, consistent with configuration theory predictions and prior evidence that, when implemented appropriately, any one of the strategic types can lead to superior performance (e.g., Conant, Mokwa, and Varadarajan 1990; Slater and Narver 1993), we checked that marketing performance variations between firms in our data set were not simply a result of differences in strategic type. We validated this assumption with analysis of variance tests that revealed no significant relationship between strategic type and either marketing performance outcome. Second, we checked that our ideal profiles correctly identified marketing organizations that contributed to superior performance by being configured in ways that fit the requirements of the business's strategic type--and not just those that were high performers regardless of their fit with strategic type. We compared the marketing performance outcomes of deviation from two different ideal marketing organization profiles, one developed from firms of the same strategic type and one developed irrespective of the firm's strategic type (e.g., Venkatraman 1990). This analysis (Table 3) validates our second assumption by indicating that calibrating ideal marketing organization profiles within strategic-type groups produces stronger deviation term coefficients and greater explanatory power in the regression models.

As shown in Table 4, the results of our hypothesis testing provide support for H₁, which predicts that the more similar a business's marketing organization profile is to that of the ideal marketing organization for its strategic type, the greater is its marketing effectiveness. Our marketing effectiveness regression models show significant, negative coefficients for deviation from the effectiveness-maximizing ideal profile for businesses implementing a prospector strategy (β = -.42, p = .008), an analyzer strategy (β = -.64, p = .0001), and a defender strategy (β = -.28, p = .02). Confidence in the power of these tests is provided in the nonideal regression models that indicate no significant relationship between deviation from the nonideal profile and marketing effectiveness for any of the three strategic types.

In H₂, we predicted that the more similar a business's marketing organization profile is to that of the ideal marketing organization for its strategic type, the greater is its marketing efficiency. This yielded mixed results. We observed significant, positive coefficients in the models that regressed deviation from the efficiency-maximizing ideal marketing organization profile against marketing efficiency in businesses implementing a prospector strategy (β = .69, p = .0002) and those pursuing a defender strategy (β = .33, p = .02). However, in analyzer strategic types, the relationship between deviation from the ideal marketing organization profile and marketing efficiency, though in the expected direction, is insignificant (β = .24, p = .17). Confidence in the power of these tests is provided in the nonideal profile regression models that indicate no significant relationships between profile deviation and marketing efficiency.

Our results indicate that organizing marketing activities in ways that fit the business's strategic type is associated with marketing effectiveness in each of the three strategic types and with marketing efficiency in firms pursuing prospector and defender strategies. This provides empirical support for strategic marketing theory predictions linking marketing organization fit with business strategy and marketing performance (e.g., Walker and Ruekert 1987). Although the total variance explained in our regression equations is moderate (ranging from 9% to 46%), these values are in line with configuration studies in the management literature (e.g., Doty, Glick, and Huber 1993; Powell 1992). The profiles of the ideal marketing organizations revealed in Appendix B are also broadly consistent with previously untested systems- structural theory propositions regarding structural differences between firms implementing different strategic types (e.g., Ruekert, Walker, and Roering 1985; Walker and Ruekert 1987).

From a strategic marketing theory perspective, these findings highlight the importance of strategy implementation as a source of competitive advantage. Our results indicate that marketing organization fit with strategic type, a key enabler of strategy implementation in marketing theory (e.g., Bonoma 1985; Walker and Ruekert 1987), is significantly associated with marketing performance. In contrast, our data suggest that a business's strategic type alone is not significantly associated with marketing performance. These findings are consistent with two central tenets of strategic marketing theory that have received only limited empirical attention: When implemented successfully, several different strategies can lead to superior performance (e.g., Day and Wensley 1988), and the way marketing activities are organized is an important enabler of strategy implementation (e.g., Walker and Ruekert 1987).

Our findings also indicate the existence of important trade-offs between the effectiveness and efficiency dimensions of marketing performance (e.g., Bonoma and Clark 1988; Walker and Ruekert 1987). For example, we observed a negative correlation between marketing efficiency and effectiveness in our data (Table 1). We found that of the 15 top performers (5 for each strategic type) used to calibrate the ideal marketing organization profiles for each dimension of marketing performance, only one firm appeared as a top performer on both marketing effectiveness and marketing efficiency performance dimensions. This highlights the need for researchers to specify and explore relationships involving different dimensions of marketing performance in empirical research (e.g., Clark 2000; Day and Wensley 1988; Slater 1995).

From a methodological perspective, our study demonstrates the utility of profile deviation approaches in assessing fit-performance relationships in strategic marketing theory. Although these approaches have been adopted in the organization theory and strategic management fields, they have not been used previously in the marketing literature. Profile deviation approaches enable researchers to assess fit in a way that is consistent with the multidimensional and holistic perspectives used in theorizing about marketing strategy. By enabling multiple variables to be assessed simultaneously, this approach also enables researchers to more closely represent the complex constructs and multiple contingencies faced by managers in the "real world" (e.g., Gresov 1989). More traditional approaches, such as moderated regression analysis, slope analysis, and subgroup analysis, can be effective in assessing fit-performance relationships involving small numbers of variables. However, these approaches are unable to effectively deal with the complex and holistic views of organization, strategy, and environment common to marketing theory (e.g., Drazin and Van de Ven 1985; Schoonhoven 1981; Venkatraman and Camillus 1984). Therefore, the profile deviation approach offers an important theory-building and theory-testing method for marketing strategy research.

From a managerial perspective, our findings highlight the need for managers to understand the multiple variables that are important characteristics of the way marketing activities are arranged and the ways they must be configured to fit the implementation requirements of the firm's business strategy. Although research on marketing organization has traditionally focused on the importance of fit between marketing organization and the served market (e.g., Achrol 1991; Day 1994), our findings indicate that to enhance performance, marketing organizations also must fit the business's strategic type (e.g., Workman, Homburg, and Gruner 1998). In designing marketing organizations to fit with business strategy, our research indicates that managers should not seek a single marketing organization template that will be both effective and efficient across different strategic types (see Aufreiter, George, and Lempres 1996; Ruekert, Walker, and Roering 1985). Rather, our findings suggest that managers should be guided by the business's strategic goals and the implementation needs of its strategic type in designing and managing their marketing organization (Walker and Ruekert 1987; Workman, Homburg, and Gruner 1998).

Furthermore, the profile deviation method used in our study and the results we obtained may be useful to managers from a benchmarking perspective. Although benchmarking has been a popular management tool in areas such as operations and quality management, its use in marketing is less common. Benchmarking involves four key stages: ( 1) identifying a firm or group of firms with superior performance, ( 2) calibrating the business processes or characteristics believed to be important in creating superior performance in the benchmark firm, ( 3) identifying gaps between the benchmark firm and the firm undertaking the benchmarking, and ( 4) developing and executing gap-closing improvement strategies to move closer to the benchmark (e.g., Camp 1989; Day 1994). The first three stages of this process are consistent with the profile deviation method outlined here. Consultants and consortia of interested firms could use this approach to undertake detailed benchmarking studies. Such studies would be able to validate the assumed link between superiority on a key business process or characteristic and superior performance and provide insights regarding the profile of the specific business processes or characteristics associated with superior performance.

For example, a trucking firm interested in benchmarking its marketing organization design could use our results as the basis of a rigorous benchmarking exercise because our study ( 1) identifies groups of firms that are particularly high performers, ( 2) calibrates their marketing organization profiles, and ( 3) demonstrates that deviation from these marketing organization profiles is associated with firm performance. Having established that marketing organization fit with business strategy is an important driver of marketing performance, managers can use the profiles of the topperforming marketing organizations in Appendix B to calibrate their own marketing organization. Managers can distinguish the appropriate benchmark for their firm by first identifying whether their firm emphasizes efficiency or effectiveness in its strategic goals and then comparing its business strategy with the strategy type descriptions in Appendix A. For example, managers in a firm that emphasizes efficiency goals with a business strategy conforming to the prospector strategy type could calibrate their marketing organization characteristics against those of similar top-performing firms in Appendix B to identify which particular marketing organization characteristics need to be changed to get closer to the benchmark profile that delivers superior marketing efficiency (see Day 1994).

As a result of trade-off decisions in research design, our study has several limitations. First, the single industry setting of our study limits the generalizability of the findings. Although such research designs are necessary to control for industry effects and isolate the fit-performance relationships of interest, studies in additional industries and multi-industry studies are needed to establish the generalizability of our findings. Second, given the novelty of the approach adopted in our study, we were conservative in our marketing organization variable selection and measurement choices to ensure that our results would be robust. Therefore, we selected only those organizational characteristics that have been highlighted as important in both configuration theory and marketing strategy studies, have well-established operationalizations to minimize measurement error, and are viewed as important by managers in our trucking industry context. Given the emergence of new virtual organizational forms (e.g., network organizations), the development of newer terms for some organizational phenomena (e.g., "empowerment"), and the need for studies in additional industries, further research may need to examine different sets of marketing organization variables.

Third, we used an empirical approach to identify ideal profiles in assessing marketing organization fit with strategic type. This is a valid and appropriate research design choice in domains in which existing knowledge is insufficient to objectively estimate theoretically derived ideal profiles. However, another intriguing possibility suggested by the management literature is the use of "experts" to derive theory-based normative ideal profiles. For example, Doty, Glick, and Huber (1993) collected questionnaire data from three of the primary researchers involved in the original development of Miles and Snow's (1978) strategic types to provide numerical estimates for theoretically derived ideal profiles of corporate characteristics appropriate for each strategic type. In exploring fit relationships in marketing strategy domains in which established theories are well accepted, future researchers may be able to similarly use data from the theory's authors as a mechanism for developing ideal profiles.

Fourth, although our study addresses the theoretically important but previously neglected question of fit between marketing organization and business strategy, we do not address the issue of the coalignment (or internal consistency) among the different marketing organization characteristics. The relationships between the multiple variables that constitute the marketing organization are a theoretically interesting and managerially difficult issue on which there has been little theoretical or empirical work. Having demonstrated the performance consequences of fit between marketing organization characteristics and business strategy, it is now important to gain an understanding of how to coalign the multiple characteristics of marketing organizations to achieve such fit. Managers need to understand how the various "levers" of marketing organization are connected to one another if they are to successfully configure marketing organizations capable of executing the firm's business strategy in ways that deliver desired strategic goals. This requires further research focused on the interrelationships between the multiple variables that are important characteristics of marketing organization.

Beyond these limitations, our results, which indicate that the relationship between marketing organization fit with strategic type and marketing performance varies across strategic types and between marketing performance dimensions, raise several questions for further research. For example, our findings suggest a particularly strong relationship between marketing organization fit with strategy type and marketing efficiency in prospector firms but weaker relationships on both dimensions of performance for defender firms. Although this could be connected with marketing having a more important role in implementing prospector strategies than defender strategies (e.g., Walker and Ruekert 1987), there is no theoretical reason that these relationships should be so varied across performance dimensions. Similarly, why does marketing organization fit with strategic type have a greater impact on marketing effectiveness in analyzers than in either prospector or defender types, but no significant impact on analyzer's marketing efficiency? The literature identifies the analyzer strategy type as difficult to execute successfully because of the conflicting demands of the simultaneous internal and external orientation required (e.g., Slater and Narver 1993). However, there is no obvious reason that this should result in such different relationships between marketing organization fit with the analyzer strategy type and the effectiveness and efficiency dimensions of marketing performance. Given the importance of these questions to managers engaged in designing and managing marketing organizations, understanding the reasons for these different relationships is a priority for further research.

Holistically framed fit-performance relationships involving strategy, organization, and environment are central to strategic marketing theory but are rarely assessed in empirical research (e.g., Day 1999; Walker and Ruekert 1987). We demonstrate that by drawing on configuration theory conceptualizations and methodological tools, many of these fit- performance relationships can be empirically assessed in ways that match their multidimensional conceptualization and holistic framing. As an example, our results indicate that organizing marketing activities in ways that fit business's strategy type can form a significant source of competitive advantage (e.g., Walker and Ruekert 1987). Given the importance of fit-performance relationships in strategic marketing theory and managers' interest in identifying such valuable sources of competitive advantage, additional studies of this type are clearly needed to enhance marketing strategy scholars' contribution to theory development and practice.

The authors gratefully acknowledge insightful comments in the development of this article from Barry Bayus, Jeff Conant, D. Harold Doty, William D. Perreault Jr., Valarie Zeithaml, and the anonymous JM reviewers.

[1] A fourth strategic type, reactors, is also identified but is deemed not to be viable in the long run as it represents firms that have no clear or consistent strategy (e.g., McKee, Varadarajan, and Pride 1989).

[2] These returns were from firms reporting a reactor strategy. Because of the small number of these respondents and the inconsistency of the reactor strategy type, we follow previous empirical studies and exclude these firms from our analysis (e.g., McDaniel and Kolari 1987; Slater and Narver 1993).

[3] The difference in the directions of the relationships is due to our marketing effectiveness measure running from low to high. Marketing efficiency is a ratio in which a smaller number represents greater efficiency.

DIAGRAM: FIGURE 1 Marketing Organization Fit with Strategic Type and Its Relationship with Marketing Performance

Legend for Chart
A = Mean
B = Standard Deviation
C = X1
D = X2
E = X3
F = X4
G = X5
H = X6
I = X7
J = X8
K = X9
L = X10
                      A      B      C      D      E     F   G      H
                      I      J      K      L

X1. Centralization
                    2.80   1.39   .84

X2. Formalization
                    4.85   1.27   .28**  .78

X3. Specialization
                    4.11   1.36  -.36**  .17*  .72

X4. Size
                    1335   3231  -.01    .01   .12    N/A

X5. Task complexity
                    4.18   1.06  -.20**  .15*  .27** .01   .79

X6. Work group
    interdependence
                    4.00   1.79  -.22**  .07   .15*  .01   .02    N/A

X7. Architectural
    marketing
    capabilities
                    3.49   1.31  -.46**  .17*  .54** .08  .29**  .28**
                     .78

X8. Specialized
    marketing
    capabilities
                    3.34   1.11  -.41**  .02   .33** .04  .26**  .13[T]
                     .64**  .72

X9. Marketing
    effectiveness
                    3.91   1.51   -.30**  .10   .30** -.07 .27**   .03
                     .50**  .60**   .85

X10. Marketing
     efficiency      .05    .16    .06   -.07   -.06  -.03 .02    -.10
                    -.10    .03   -.13[T]  N/A

[T]p < .10.

*p < .05.

**p < .01.

Notes: Alphas are shown on the correlation matrix diagonal. N/A = not applicable.

  CME           GM           Mean
 Rater         Rater      Inter-Rater     t-Value      Inter-Rater
Mean (S.D.)  Mean (S.D.)  Difference[a] (significance) Correlation[b]

Centralization
2.57 (1.51)    2.42 (1.14)     .15          .67 (.51)         .65


Formalization
5.10 (1.37)    5.44 (.99)     -.34          .95 (.35)         .56

Specialization
4.58 (1.62)    4.43 (1.19)     .15          .39 (.70)         .56

Task complexity
4.70 (.92)     4.97 (1.28)    -.27         1.10 (.29)         .71

Work group interdependence
4.00 (1.60)    3.42 (2.04)     .58          .86 (.40)         .58

Specialized marketing capabilities
4.97 (1.05)    5.06 (1.15)    -.09          .28 (.78)         .55

Architectural marketing capabilities
4.77 (1.19)    5.14 (1.10)    -.37         1.53 (.14)         .60

Marketing effectiveness
5.26 (1.40)    5.53 (1.49)    -.27          .67 (.51)         .57

[a]Inter-rater difference is CME mean score less GM mean score.

[b]All correlations significant at p < .001 level.

Notes: S.D. = standard deviation.

Legend for Chart

A = Independent Variables
B = Dependent Variable Marketing Effectiveness Within Strategic-
    Type Group Ideal Profile Model
C = Dependent Variable Marketing Effectiveness Across Strategic-
    Type Group Ideal Profile Model
D = Dependent Variable Marketing Efficiency Within Strategic-
    Type Group Ideal Profile Model
E = Dependent Variable Marketing Efficiency Across Strategic-
    Type Group Ideal Profile Model

   A                     B          C          D         E

All Firms

  Profile deviation    -.44**     -.39**      .29**     .12

  Firm size (log)       .09       -.02        .18*      .22*

  R[2]                  .20        .15        .13       .06

  F-value             19.49**    14.09**     8.39**    3.96**

*p < .05.

**p < .01.

TABLE 4 Marketing Organization Fit with Strategic Type and Performance Regression Models

Legend for Chart

A = Independent Variables
B = Dependent Variable Marketing Effectiveness
    Ideal Profile Models
C = Dependent Variable Marketing Effectiveness
    Nonideal Models
D = Dependent Variable Marketing Efficiency
    Ideal Profile Models
E = Dependent Variable Marketing Efficiency
    Nonideal Models

   A                     B          C          D         E

Prospectors

 Profile deviation    -.42**     -.01        .69**      .22

 Organization size
 (log)                 .23*       .28       -.15       -.10

 R[2]                  .26        .08        .46        .07

 F-value              6.12**     1.43       9.70**      .92


Analyzers

 Profile deviation    -.64**      .21         .24      -.19

 Organization size
 (log)                 .18       -.04         .43*      .40*

 R[2]                  .41        .04         .15       .12

 F-value             18.08**     1.24        3.20*     2.78


Defenders

 Profile deviation    -.28*       .15         .33*      .16

 Organization size
 (log)                -.11       -.16         .22       .27*

 R[2]                  .09        .05         .15       .10

 F-value              3.20*      1.55        3.98*     2.58

*p < .05.

**p < .01.

Centralization (seven-point scale with "strongly disagree" and "strongly agree" as anchors)

Source: Jaworski and Kohli (1993) The following questions concern how decisions are made in your marketing organization.

How strongly do you agree or disagree with each of the following statements about your marketing organization?

• There can be little action taken in the marketing organization until a supervisor makes a decision.
• A person who wants to make his or her own decisions would be quickly discouraged in the marketing organization.
• Even small matters have to be referred to someone with more authority for a final decision.
• Any decision a person in the marketing organization makes has to have his or her boss's approval.

Formalization (seven-point scale with "strongly disagree" and "strongly agree" as anchors)

Source: Deshpande and Zaltman (1982)

The following questions concern the impact of work rules used in your marketing organization.

How strongly do you agree or disagree with each of the following statements about your marketing organization?

• Most people in the marketing organization follow written work rules for their job.
• How things are done in the marketing organization is never left up to the person doing the work.
• People in the marketing organization are allowed to do almost as they please when performing their work. (RS)

Specialization (seven-point scale with "strongly disagree" and "strongly agree" as anchors)

Source: Doty, Glick, and Huber (1993)

The following questions concern job responsibilities and skills within your marketing organization.

How strongly do you agree or disagree with each of the following statements about your marketing organization?

• Marketing personnel in this firm have very specific job responsibilities.
• Most marketing employees have jobs that require special skills.
• Standardized training procedures exist for marketing jobs. (RS)
• Written position descriptions are provided to marketing specialists.

Specialized Marketing Capabilities (seven-point scale with "not very well" and "very well" as anchors)

Source: New Scale

How well does your organization perform the following activities relative to competitors...

• advertising and promotion
• public relations
• personal selling
• new product/service development

Architectural Marketing Capabilities (seven-point scale with "not very well" and "very well" as anchors)

Source: New Scale

How well does your organization perform the following activities relative to competitors...

• environmental scanning
• market planning
• marketing skill development
• marketing implementation

Task Complexity (seven-point scale with "not at all" and "to a great extent" as anchors)

Source: Doty, Glick, and Huber (1993)

To what extent...

• is the work that people in the marketing organization do the same from day to day?
• does the work move among the marketing work groups in a sequential manner?
• is there a clearly known way to do the major types of work that marketing work groups deal with?
• do marketing employees tend to perform the same tasks in the same way?
• is there an understandable sequence of steps that can be followed to perform most marketing tasks?

Work Group Interdependence (seven-point scale with "not at all" and "to a great extent" as anchors)

Source: Van de Ven, Delbecq, and Koenig (1976)

To what extent does the flow of work in the department reflect the diagram below, in which work comes into the department and different subunits diagnose, problem solve, and work together as a group at the same time?

Materials, customer orders, and information enter the marketing
department.

                     (Down arrows)

Same Time       Subunit 1      Subunit 2
                Subunit 3      Subunit 4

                      (Down arrows)

Services and information leave the marketing department.

Strategic Types

Source: McKee, Varadarajan, and Pride (1989)

The following descriptions characterize equally effective strategies that organizations can use to position themselves relative to their competition. Please select the description that you feel best characterizes your firm today.

Prospector Strategy: This business unit typically operates within a broad product-market domain that undergoes periodic redefinition. The business unit values being "first in" in new product and market areas even if not all these efforts prove to be highly profitable. This organization responds rapidly to early signals concerning areas of opportunity, and these responses often lead to a new round of competitive actions. However, this business unit may not maintain market strength in all areas it enters.

Analyzer Strategy: This business unit attempts to maintain a stable, limited line of products or services while moving quickly to follow a carefully selected set of the more promising new developments in the industry. This organization is seldom "first in" with new products and services. However, by carefully monitoring the actions of major competitors in areas compatible with its stable product-market base, this business unit can frequently be "second in" with a more cost-efficient product or service.

Defender Strategy: This business unit attempts to locate and maintain a secure niche in a relatively stable product or service area. The business unit tends to offer a more limited range of products or services than competitors, and it tries to protect its domain by offering higher quality, superior service, lower prices, and so forth. Often, this business unit is not at the forefront of developments in the industry. It tends to ignore industry changes that have no direct influence on current areas of operation and concentrates instead on doing the best job possible in a limited area.

Marketing Effectiveness (seven-point scale with "not very well" and "very well" as anchors) How well has your firm achieved its goals in terms of ...

• market share growth
• sales growth
• market position

Marketing Efficiency (objective data from TTS database)

Marketing and selling expenses/gross revenue

Notes: (RS) = reverse scoring.

Legend for Chart
A = Marketing Organization Characteristics
B = Prospector Firms Marketing Effectiveness
C = Prospector Firms Marketing Efficiency
D = Analyzer Firms Marketing Effectiveness
E = Analyzer Firms Marketing Efficiency
F = Defender Firms Marketing Effectiveness
G = Defender Firms Marketing Efficiency

A                 B      C      D      E      F      G

Centralization   2.25   2.30   2.10   2.10   3.50   3.75

Formalization    4.67   4.33   5.47   5.53   6.00   5.28

Specialization   5.75   4.45   5.50   4.65   4.15   3.33

Task complexity  4.04   3.84   5.24   4.00   5.12   4.47

Work group
interdependence  3.00   4.00   4.40   3.00   4.00   3.33

Specialized
marketing
capabilities     4.24   3.70   5.62   3.53   4.00   3.62

Architectural
marketing
capabilities     4.76   4.34   5.38   3.81   4.44   3.25

REFERENCES

Achrol, Ravi S. (1991), "Evolution of the Marketing Organization: New Forms for Turbulent Environments," Journal of Marketing, 55 (October), 77-93.

Aiken, Michael and Jerald Hage (1968), "Organizational Interdependence and Intra-organizational Structure," American Sociological Review, 33 (6), 912-30.

American Trucking Association (1999), Standard Trucking and Transportation Statistics. Alexandria, VA: ATA Economics & Statistical Analysis Department.

Anderson, James C. and David W. Gerbing (1988), "Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach," Psychological Bulletin, 103 (3), 411-23.

Armstrong, J. Scott and Terry S. Overton (1977), "Estimating Non-response Bias in Mail Surveys," Journal of Marketing Research, 14 (August), 396-402.

Aufreiter, Nora, Mike George, and Liz Lempres (1996), "Developing a Distinctive Consumer Marketing Organization," Journal of Market-Focused Management, 1 (3), 199-208

Bagozzi, Richard P. and Lynn W. Phillips (1982), "Representing and Testing Organizational Theories: A Holistic Construal," Administrative Science Quarterly, 27 (September), 459-89.

Barney, Jay B. (1991), "Firm Resources and Sustained Competitive Advantage," Journal of Management, 17 (March), 99-120.

Bentler, Peter M. and Chih-Ping Chou (1987), "Practical Issues in Structural Modeling," Sociological Methods and Research, 16 (1), 78-117.

Bharadwaj, Sundar G., P. Rajan Varadarajan, and John Fahy (1993), "Sustainable Competitive Advantage in Service Industries: A Conceptual Model and Research Propositions," Journal of Marketing, 57 (October), 83-99.

Bhargava, Mukesh, Chris Dubelaar, and Sridhar Ramaswami (1994), "Reconciling Diverse Measures of Performance: A Conceptual Framework and a Test of Methodology," Journal of Business Research, 31 (5), 235-46.

Bonoma, Thomas V. (1985), The Marketing Edge: Making Strategies Work. New York: The Free Press.

------ and Bruce H. Clark (1988), Marketing Performance Assessment. Cambridge, MA: Harvard Business School Press.

Boyer, Kenneth D. (1993), "Deregulation of the Trucking Sector: Specialization, Concentration, Entry, and Financial Distress," Southern Economic Journal, 59 (3), 481-94.

Camp, Robert C. (1989), Benchmarking: The Search for Best Practices that Lead to Superior Performance. Milwaukee, WI: ASQC Quality Press.

Cespedes, Frank V. (1991), Organizing and Implementing the Marketing Effort. Cambridge, MA: Addison-Wesley.

------ (1995), Concurrent Marketing: Integrating Product, Sales, and Service. Cambridge, MA: Harvard Business School Press.

Clark, Bruce H. (2000), "Managerial Perceptions of Marketing Performance: Efficiency, Adaptability, Effectiveness, and Satisfaction," Journal of Strategic Marketing, 8 (1), 3-26.

Conant, Jeffrey S., Michael P. Mokwa, and P. Rajan Varadarajan

(1990), "Strategic Types, Distinctive Marketing Competencies, and Organizational Performance: A Multiple Measures-Based Study," Strategic Management Journal, 11 (5), 365-83.

Day, George S. (1994), "The Capabilities of Market-Driven Organizations," Journal of Marketing, 58 (October), 37-52.

------ (1997), "Aligning the Organization to the Market," in Reflections on the Futures of Marketing, Donald R. Lehmann and Katherine E. Jocz, eds. Cambridge, MA: Marketing Science Institute.

------ (1999), Market-Driven Organization. Cambridge, MA: The Free Press.

------ and Robin Wensley (1988), "Assessing Advantage: A Framework for Diagnosing Competitive Superiority," Journal of Marketing, 52 (April), 1-20.

Deshpande, Rohit and Gerald Zaltman (1982), "Factors Affecting the Use of Market Research Information: A Path Analysis," Journal of Marketing Research, 19 (February), 14-31.

Dess, Gregory G., Stephanie Newport, and Abdul M. A. Rasheed (1993), "Configuration Research in Strategic Management: Key Issues and Suggestions," Journal of Management, 19(4), 775-95.

Doty, D. Harold, William H. Glick, and George P. Huber (1993), "Fit, Equifinality, and Organizational Effectiveness: A Test of Two Configurational Theories," Academy of Management Journal, 30 (December), 1196-1250.

Drazin, Robert and Andrew H. Van de Ven (1985), "Alternative Forms of Fit in Contingency Theory," Administrative Science Quarterly, 30 (December), 514-39.

Germain, Richard, Cornelia Droge, and Patricia Daugherty (1994), "The Effect of Just-in-Time Selling on Organizational Structure: An Empirical Investigation," Journal of Marketing Research, 31 (November), 471-83.

Grant, Robert M. (1996), "Prospering in Dynamically-Competitive Environments: Organizational Capability as Knowledge Integration," Organization Science, 7 (July-August), 375-87.

Gresov, Christopher (1989), "Exploring Fit and Misfit with Multiple Contingencies," Administrative Science Quarterly, 34 (September), 431-53.

Hughes, Marie Adele and Dennis E. Garrett (1990), "Intercoder Reliability Estimation Approaches in Marketing: A Generalizability Framework for Quantitative Data," Journal of Marketing Research, 27 (May), 185-95.

Inkson, J.H.K., Derek Pugh, and D. Hickson (1970), "Organization, Context, and Structure: An Abbreviated Replication," Administrative Science Quarterly, 15 (June), 318-29.

James, William L. and Kenneth J. Hatten (1995), "Further Evidence on the Validity of the Self-Typing Paragraph Approach: Miles and Snow Strategic Archetypes in Banking," Strategic Management Journal, 16 (October), 161-68.

Jaworski, Bernard J. and Ajay K. Kohli (1993), "Market Orientation: Antecedents and Consequences," Journal of Marketing, 57 (July), 53-70.

Jennings, Daniel F. and Samuel L. Seaman (1994), "High and Low Levels of Organizational Adaptation: An Empirical Analysis of Strategy, Structure, and Performance," Strategic Management Journal, 15 (July), 459-75.

John, George and Claude Martin (1984), "Effects of Organization Structure on Marketing Planning and Credibility and Utilization of Plan Output," Journal of Marketing Research, 21 (May), 170-83.

Ketchen, David J., James G. Combs, Craig J. Russell, Chris Shook,

Michelle A. Dean, Janet Runge, Franz T. Lohrke, Stefanie Nauman, Dawn Ebe Haptonstahl, Robert Baker, Brendan A. Beck-stein, Charles Handler, Heather Honig, and Stephen Lamoureux (1997), "Organizational Configurations and Performance: A Meta-Analysis," Academy of Management Journal, 40 (1), 223-40.

------, James B. Thomas, and Charles C. Snow (1993), "Organizational Configurations and Performance: A Comparison of Theoretical Approaches," Academy of Management Journal, 36 (6), 1278-1313.

Kohli, Ajay K. and Bernard J. Jaworski, (1994), "The Influence of Coworker Feedback on Salespeople," Journal of Marketing, 58 (October), 82-84.

Lambert, Douglas M., Christine Lewis, and James R. Stock (1993), "Customer-Focused Strategies for Motor Carriers," Transportation Journal, 32 (4), 21-28.

Lewin, Arie Y. and John W. Minton (1986), "Determining Organizational Effectiveness: Another Look and an Agenda for Research," Management Science, 32 (May), 514-38.

MacLeod, Mary D., Lawrence L. Garber, Michael J. Dotson, and Terry M. Chambers (1999), "The Use of Promotional Tools in the Motor Carrier Industry: An Exploratory Study," Transportation Journal, 38 (3), 42-56.

Matsuno, Ken and John T. Mentzer (2000), "The Effects of Strategy Type on the Market Orientation-Performance Relation-ship," Journal of Marketing, 64 (October), 1-16.

McDaniel, Stephen W. and James W. Kolari (1987), "Marketing Strategy Implications of the Miles and Snow Strategic Typology," Journal of Marketing, 51 (October), 19-30.

McKee, Daryl O., P. Rajan Varadarajan, and William M. Pride (1989), "Strategic Adaptability and Firm Performance: A Market-Contingent Perspective," Journal of Marketing, 53 (July), 21-35.

Menon, Anil and P. Rajan Varadarajan (1992), "A Model of Marketing Knowledge Use Within Firms," Journal of Marketing, 56 (October), 53-71.

Meyer, Alan D., Anne S. Tsui, and C.R. Hinings (1993), "Configurational Approaches to Organizational Analysis," Academy of Management Journal, 30 (December), 1175-95.

Miles, Raymond E. and Charles C. Snow (1978), Organizational Strategy, Structure, and Process. New York: McGraw-Hill.

------ and ------ (1994), Fit, Failure, and the Hall of Fame. New York: Macmillan.

Milgrom, Paul and John Roberts (1995), "Complementarities and Fit: Strategy, Structure, and Organizational Changes in Manufacturing," Journal of Accounting and Economics, 19 (March- May), 179-208.

Miller, Danny (1997), "Configurations Revisited," Strategic Management Journal, 13 (Summer), 111-25.

------ and Henry Mintzberg (1988), "The Case for Configuration," in The Strategy Process: Concepts, Contexts, and Cases, James B. Quinn and Robert M. James, eds. Englewood Cliffs, NJ: Prentice Hall.

Moorman, Christine, Rohit Deshpande, and Gerald Zaltman (1993), "Factors Affecting Trust in Market Research Relationships," Journal of Marketing, 57 (January), 81-101.

------ and Anne S. Miner (1997), "The Impact of Organizational Memory on New Product Performance and Creativity," Journal of Marketing Research, 34 (February), 91-106.

------ and Roland T. Rust (1999), "The Role of Marketing," Journal of Marketing, 63 (Special Issue), 180-97.

Morgan, Neil A., Bruce H. Clark, and Rich Gooner (2002), "Marketing Productivity, Marketing Audits, and Systems for Marketing Performance Assessment: Integrating Multiple Perspectives," Journal of Business Research, 55 (5), 363-75.

Noble, Charles H. and Michael P. Mokwa (1999), "Implementing Marketing Strategies: Developing and Testing a Managerial Theory," Journal of Marketing, 63 (October), 57-73.

Olson, Eric M., Orville C. Walker, and Robert W. Ruekert (1995), "Organizing for Effective New Product Development: The Moderating Role of Product Innovativeness," Journal of Marketing, 59 (January), 48-62.

Ostroff, Cheri and Neal Schmitt (1993), "Configurations of Organizational Effectiveness and Efficiency," Academy of Management Journal, 36 (6), 1345-61.

Porter, Michael E. (1996), "What Is Strategy?" Harvard Business Review, 74 (November-December), 61-78.

Powell, Thomas C. (1992), "Organizational Alignment as Competitive Advantage," Strategic Management Journal, 13 (September), 119-34.

Ruekert, Robert W. and Orville C. Walker (1987), "Interactions Between Marketing and R&D Departments in Implementing Different Business Strategies," Strategic Management Journal, 8 (May-June), 233-48.

------, ------, and Kenneth J. Roering (1985), "The Organization of Marketing Activities: A Contingency Theory of Structure and Performance," Journal of Marketing, 49 (Winter), 13-25.

Schoonhoven, Claudia B. (1981), "Problems with Contingency Theory: Testing Assumptions Hidden Within the Language of Contingency Theory," Administrative Science Quarterly, 26 (June), 349-77.

Shortell, Steven M. and Edward J. Zajac (1990), "Perceptual and Archival Measures of Miles and Snow's Strategic Types: A Comprehensive Assessment of Reliability and Validity," Academy of Management Journal, 33 (4), 817-32.

Silverman, Brian S., Jack A. Nickerson, and John Freeman (1997), "Profitability, Transactional Alignment, and Organizational Mortality in the U.S. Trucking Industry," Strategic Management Journal, 18 (Summer), 31-52.

Slater, Stanley F. (1995), "Issues in Conducting Marketing Strategy Research," Journal of Strategic Marketing, 3(December), 257-70.

------ and John C. Narver (1993), "Product-Market Strategy and Performance: An Analysis of the Miles and Snow Strategic Types," European Journal of Marketing, 27 (10), 33-51.

------ and Eric M. Olson (2000), "Strategy Type and Performance: The Influence of Sales Force Management," Strategic Management Journal, 21 (August), 813-29.

Tsui, Anne S. (1990), "A Multiple-Constituency Model of Effectiveness: An Empirical Examination at the Human Resource Subunit Level," Administrative Science Quarterly, 35, 458-83.

U.S. Bureau of the Census (1998), Transportation Annual Survey. Washington, DC: U.S. Department of Commerce.

Van de Ven, Andrew H., André L. Delbecq, and Richard Koenig (1976), "Determinants of Coordination Mode Within Organizations," American Sociological Review, 41 (April), 322-38.

------ and Robert Drazin (1985), "The Concept of Fit in Contingency Theory," in Research in Organizational Behavior, Vol. 7, Larry L. Cummings and Barry M. Staw, eds. Greenwich, CT: JAI Press, 333-65.

Veliyath, Rajaram and T.C. Srinivasan (1995), "Gestalt Approaches to Assessing Strategic Coalignment: A Conceptual Integration," British Journal of Management,6 (September), 205-19.

Venkatraman, N. (1989), "The Concept of Fit in Strategy Research: Toward Verbal and Statistical Correspondence," Academy of Management Review, 14 (3), 423-44.

------ (1990), "Performance Implications of Strategic Coalignment: A Methodological Perspective," Journal of Management Studies, 27 (1), 19-41.

------ and John C. Camillus (1984), "Exploring the Concept of 'Fit' in Strategic Management," Academy of Management Review, 9 (4), 513-25.

------ and John E. Prescott (1990), "Environment-Strategy Coalignment: An Empirical Test of Its Performance Implications," Strategic Management Journal, 11 (January), 1-23.

------ and V. Ramanujam (1986), "Measurement of Business Performance in Strategy Research: A Comparison of Approaches," Academy of Management Review, 11 (October), 801-14.

Walker, Orville C. (1997), "The Adaptability of Network Organizations: Some Unexplored Questions," Journal of the Academy of Marketing Science, 25 (1), 75-82.

------ and Robert W. Ruekert (1987), "Marketing's Role in the Implementation of Business Strategies: A Critical Review and Conceptual Framework," Journal of Marketing, 51(July), 15-33.

Webster, Frederick (1997), "The Future Role of Marketing in the Organization," in Reflections on the Futures of Marketing, Donald R. Lehmann and Katherine E. Jocz, eds. Cambridge, MA: Marketing Science Institute.

Workman, John P., Christian Homburg, and Kjell Gruner (1998), "Marketing Organization: An Integrative Framework of Dimensions and Determinants," Journal of Marketing, 62 (July), 21-41.

Zahra, Shaker A. and John A. Pearce (1990), "Research Evidence on the Miles and Snow Typology," Journal of Management, 16 (December), 751-68.

Zajac, Edward J., Mathew S. Kraatz, and Rudi K.F. Bresser (2000), "Modeling the Dynamics of Strategic Fit: A Normative Approach to Strategic Change," Strategic Management Journal, 21 (4), 429-53.

~~~~~~~~

By Douglas W. Vorhies and Neil A. Morgan

Douglas W. Vorhies is Assistant Professor of Marketing, College of Business, Illinois State University. Neil A. Morgan is Assistant Professor of Marketing, Kenan-Flagler Business School, University of North Carolina at Chapel Hill.

A Configurational Perspective on Key Account Management 

Most firms struggle with the challenge of managing their key customer accounts. There is a significant gap between the importance of this organizational design problem in practice and the research attention paid to it. Sound academic research on key account management (KAM) is limited and fragmented. Drawing on research on KAM and team selling, the authors develop an integrative conceptualization of KAM and define key constructs in four areas: ( 1) activities, ( 2) actors, ( 3) resources, and ( 4) approach formalization. Adopting a configurational perspective to organizational research, the authors then use numerical taxonomy to empirically identify eight prototypical KAM approaches on the basis of a cross-industry, cross-national study. The results show significant performance differences among the approaches. Overall, the article builds a bridge between marketing organization research and relationship marketing research.

Many companies today are faced with powerful and more demanding customers. In many industries, these powerful buyers have been shaped through corporate mergers and have been visible in many industry sectors such as retailing, automotive, computers, and pharmaceuticals. These large customers often rationalize their supply base to cooperate more closely with a limited number of preferred suppliers (e.g., Dorsch, Swanson, and Kelley 1998; Stump 1995). They may demand special value-adding activities from their suppliers, such as joint product development, financing services, or consulting services (Cardozo, Shipp, and Roering 1992). Also, many buying firms have centralized their procurement and expect a similarly coordinated selling approach from their suppliers. For example, global industrial customers may demand uniform pricing terms, logistics, and service standards on a worldwide basis from their suppliers (Montgomery and Yip 2000).

These demands from important accounts raise an organizational design problem for many suppliers. As Kempeners and van der Hart (1999, p. 312) note, "Organizational structure is perhaps the most interesting and controversial part of account management." Internal organizational structures often hamper a coordinated account management, such as when the same customer is served by decentralized product divisions or by highly independent local sales operations. In addition, the set of activities for complex customers cannot be handled by the sales function alone but requires participation from other functional groups. These developments have induced many suppliers to rethink how they manage their most important customers and how they design their internal organization in order to be responsive to these key customers. In this context, firms are increasingly organizing around customers and shifting resources from product divisions or regional divisions to customer-focused business units (Homburg, Workman, and Jensen 2000). Many firms are establishing specialized key account managers and are forming customer teams that are composed of people from sales, marketing, finance, logistics, quality, and other functional groups (Millman 1996; Wotruba and Castleberry 1993).

In a recent study, Homburg, Workman, and Jensen (2000) argue that the increasing emphasis on key account management (KAM) is one of the most fundamental changes in marketing organization. Given the relevance of designing KAM in practice, sound academic research on this topic is still surprisingly limited. Millman (1996, p. 631) notes that "Key account management is underresearched and its efficacy, therefore, is only partially understood." Although management approaches to the most important customers have received some research attention over the past 25 years (Shapiro and Moriarty 1984a; Weilbaker and Weeks 1997), the existing literature has several shortcomings. First, research has been fragmented and has not consolidated specific design aspects of KAM into a coherent framework. Second, conceptual and empirical work on KAM has primarily been based on observations of formalized key account programs in Fortune-500 companies and has hardly been extended to nonformalized KAM approaches. Third, broad-based empirical research on KAM is still scarce, as Kempeners and van der Hart (1999, p. 311) note: "Although Stevenson (1980) noted almost 20 years ago that: 'despite widespread industrial use, there has been little empirical research on national account marketing,' it seems that this is still true." The empirical work that has been done in the past has essentially been descriptive. Finally, given that conceptual work has suggested a variety of design options (Shapiro and Moriarty 1984a), there is little empirical knowledge of which types of approaches to KAM occur in practice and how successful these are.

Given these gaps in knowledge about KAM, the overall objective of this article is to study the design of approaches to KAM. More specifically, we seek to

1. Derive the core design dimensions of KAM approaches from the KAM literature and from related research areas to develop an integrative conceptualization of KAM,

• 2. Identify the key constructs within these design dimensions and develop instruments for measuring these constructs,
• 3. Identify prototypical approaches to KAM in practice on the basis of a cross-national, cross-industry taxonomy, and
• 4. Explore the outcomes of different KAM approaches.

Given that taxonomies are less frequently developed than conceptual models, a few comments on their value are in order. As Hunt (1991, p. 176) has noted, classification schemata, such as typologies or taxonomies, "play fundamental roles in the development of a discipline since they are the primary means for organizing phenomena into classes or groups that are amenable to systematic investigation and theory development." Given that the conceptual knowledge about the design of KAM is at an early stage and that our research endeavor is to expand its scope, a taxonomy is particularly useful in providing the field with new organization. By means of the taxonomy, we are studying the complex KAM phenomenon through holistic patterns of multiple variables rather than isolated variables and their bivariate relations. This research approach is consistent with the configurational perspective of organizational analysis that has been gaining increasing attention (Meyer, Tsui, and Hinings 1993). The basic premise of the configurational perspective is that "Organizational structures and management systems are best understood in terms of overall patterns rather than in terms of analyses of narrowly drawn sets of organizational properties" (Meyer, Tsui, and Hinings 1993, p. 1181). Thus, the configurational perspective complements the traditional contingency approach (Mahajan and Churchill1990). Two alternatives of identifying configurations have been distinguished: Typologies represent classifications based on a priori conceptual distinctions, whereas taxonomies are empirically derived groupings (Hunt 1991; Rich 1992; Sanchez 1993). Given our goal of identifying approaches to KAM in practice, we take a taxonomic approach. Hunt(1991) notes that grouping phenomena through taxonomies as opposed to typologies requires substantially less a priori knowledge about which specific properties are likely to be powerful for classification, because taxonomic procedures are better equipped to handle large numbers of properties

The article is organized as follows: We first summarize the literature on KAM and evaluate contributions that the personal selling and sales management literature provide for KAM. On the basis of the literature review, we develop a multidimensional conceptualization of KAM and identify outcomes of KAM. We then describe a large-scale survey of KAM approaches and develop the taxonomy. This is followed by an exploration of how the different approaches perform. We conclude by discussing implications for theory and managerial practice.

KAM Research

We subsume under KAM all approaches to managing the most important customers that have been discussed under such diverse terms as key account selling, national account management, national account selling, strategic account management, major account management, and global account management. "National account management" has become a misnomer, as business with important customers increasingly spans country borders (Colletti and Tubridy 1987). Although some research has focused on global accounts (Montgomery and Yip 2000; Yip and Madsen 1996), KAM appears to be the most accepted term in recent publications (Jolson 1997; McDonald, Millman, and Rogers 1997; Pardo 1997; Sharma 1997) and is the most widely used term in Europe.[ 1]

Table 1 presents a summary of selected KAM research. We segment this research into articles focusing on ( 1) individual key account managers, ( 2) dyadic relationships between suppliers and key accounts, and ( 3) the design of key account programs. Given our objective of understanding the design of KAM approaches, Group 3 is most relevant to our article.

Because Group 1 takes the individual key account manager as the unit of analysis, it is similar to personal selling research. Weeks and Stevens (1997) find considerable dissatisfaction of key account managers with their current training programs. Boles, Barksdale, and Johnson (1996) identify behaviors required of key account salespeople in order to build successful key account relationships.

Group 2 is closely related to relationship marketing research. Several authors describe an evolutionary path of key account relationships from lower to higher degrees of involvement and collaboration (Lambe and Spekman 1997; McDonald, Millman, and Rogers 1997). Sharma (1997) finds that customers' preference for being served by key account programs is particularly high when their buying process is long and complex. Sengupta, Krapfel, and Pusateri (1997b) study switching costs in key account relationships.

Group 3, which focuses on overall management of key accounts, is the largest group, consistent with Pardo's (1999, p. 286) conclusion that "Today, key account experts on both sides of the Atlantic agree on ... the problem of key account management as being an organizational one." Although all studies in Group 3 deal with the design of key account programs, none of these integrates the main aspects of key account program design within one study.

Four main themes emerge from the literature on key account programs. First, key account programs encompass special (interorganizational) activities for key accounts that are not offered to average accounts. These special activities pertain to such areas as pricing, products, services, distribution, and information sharing (Cardozo, Shipp, and Roering 1992; Montgomery and Yip 2000; Shapiro and Moriarty 1984b). Second, key account programs frequently involve special (intraorganizational) actors who are dedicated to key accounts. These key account managers are typically responsible for several key accounts and report high in the organization (Colletti and Tubridy 1987; Dishman and Nitse 1998; Wotruba and Castleberry 1993). They may be placed in the supplier's headquarters, in the local sales organization of the key account's country, or even in the key account's facilities (Millman 1996; Yip and Madsen 1996). It is frequently stressed that key account managers need special compensation arrangements and skills, which has implications for their selection, training, and career paths (Colletti and Tubridy 1987; Tice 1997). Third, KAM is a multifunctional effort involving, in addition to marketing and sales, functional groups such as manufacturing, research and development, and finance (Shapiro and Moriarty 1984b). Fourth, the formation of key account programs is influenced by characteristics of buyers and of the market environment, such as purchasing centralization, purchasing complexity, demand concentration, and competitive intensity (Boles, Johnston, and Gardner 1999; Stevenson 1980).

We observe several shortcomings in prior research. First, the previous design issues have mostly been studied in isolation and have not been consolidated into a coherent framework. Shapiro and Moriarty's (1984a, p. 34) assessment that "the term national account management program is fraught with ambiguity" is still valid. Second, there is a general lack of quantitative empirical studies on the design issues, particularly on the cross-functional linkages of KAM. The quantitative research that has been undertaken has essentially been descriptive and has not systematically developed and validated measures. Third, much of the empirical work that has been done (and has driven conceptual ideas) is based on observations in large Fortune-500 companies with sophisticated, formalized key account programs. This excludes small and medium-sized companies that actively manage relationships with key accounts but do not formalize the KAM approach. Quantitative empirical research has not taken up Shapiro and Moriarty's (1984a, p. 5) comment in their early conceptual work that "the simplest structural option is no program at all." Fourth, given that conceptual work has mentioned a variety of structural options (Shapiro and Moriarty 1984a), there is no broad-based empirical work that allows generalizations about how KAM is done in practice. We now position KAM research in a wider research context and evaluate the contribution of related research to the open issues in the KAM literature.

Research Related to KAM

Key account management can be subsumed under the wider context of personal selling and sales management research. From a sales management perspective, KAM represents one element within a differentiated sales force that stands next to other elements such as telemarketing, demonstration centers, and traditional personal and face-to-face selling (Cardozo, Shipp, and Roering 1987; Marshall, Moncrief, and Lassk 1999). According to Shapiro and Wyman (1981, p. 104), "National account management thus is an extension, improvement, and outgrowth of personal selling."

Most personal selling research has a different level of analysis than our work does. Although this literature has examined relationship-building activities for important customers (Jolson 1997; Weitz and Bradford 1999; Wotruba 1991) and has produced empirical classifications based on activities (Moncrief 1986), its level of analysis is the individual salesperson. Thus, although potentially enhancing knowledge about individual key account managers, this research contributes little to understanding organizational approaches to KAM.

In recent years, however, there has been a shift in the level of analysis from the individual salesperson to the selling team (Weitz and Bradford 1999). There is growing recognition that functional groups other than sales play an important role in interactions with customers (Hutt, Johnston, and Ronchetto 1985; Spekman and Johnston 1986). The team-selling literature has distinguished between "core selling teams" that are permanently assigned to customer accounts and the wider "selling center" that consists of members of all functional groups who participate on an ad hoc basis (Moon and Gupta 1997; Smith and Barclay1990). Moon and Armstrong(1994, p. 19) explicitly link team-selling literature to KAM by noting that "conceptually, national account teams can be viewed as selling teams ... that service large, complex customers."

The team-selling literature enhances our conceptual understanding of cross-functional cooperation for key accounts. One fundamental problem for sales managers is to obtain the cooperation of other organizational members without having formal authority over them (Spekman and Johnston 1986). Therefore, the achievement of selling tasks is hypothesized to be dependent on the selling center participants' commitment to the selling team and its goals (Smith and Barclay 1993) and on their connection through communication flows (Moon and Gupta 1997). However, empirical research on team selling is just as scarce as empirical research on cross-functional cooperation in KAM.

At this point, it is important to clarify how our research perspective differs from the vast body of research on relation-ship marketing and market orientation. Relationship marketing research focuses more on interorganizational issues between suppliers and their customers, such as how marketing relationships are built and maintained and what benefits accrue (Morgan and Hunt 1999).These are mostly assessed from the customer's perspective. On the contrary, our focus is more on how firms organize and cooperate internally. In addition, our level of analysis is the KAM approach (which encompasses relationships with several important customers), whereas the unit of analysis in most of the relationship marketing literature is a given dyadic relationship with an individual customer. Because most firms have the challenge to array their organizational resources at a set of strategically important customers rather than just one customer, ours is an important perspective for study. Market orientation research, in turn, studies both intraorganizational and interorganizational cooperation to create superior value for buyers. However, this research studies constructs on a high level of abstraction. Another key difference from KAM is that market orientation literature treats the customer base as a whole and does not differentiate between important customers and average customers.

Approach to the Conceptualization

In this section, we blend the insights from prior literature into an integrative conceptualization of KAM. Our conceptualization is composed of fundamental dimensions of KAM, each of which comprises several key constructs. Because we use these constructs to develop a taxonomy of KAM approaches subsequently, we give great care to their selection. As Bailey (1994, p.2) notes, "One basic secret to successful classification, then, is the ability to ascertain the key or fundamental characteristics on which the classification is to be based." The literature suggests several different, partly contradictory guidelines for the selection of input variables to a classification (for a review, see Rich 1992). There is consensus that the input variables should be derived from theory and should be meaningful for the subject under study. Therefore, given our integrative perspective on KAM, we derive theory-based constructs from the literature that are comparable across a range of industries.

The degree of admissible interdependencies among the cluster variables is more debated. Whereas Sneath and Sokal (1973) advocate to exclude variables that are logically or empirically correlated, Arabie and Hubert (1994, p. 166) note that "it is difficult to imagine empirical data arising in the behavioral sciences that would have all columns mutually independent." In addition, from a methodological vantage point, there is no assumption of uncorrelated variables in most cluster methods (Milligan 1996, p. 347). We concur with the latter viewpoint in that we accept some conceptual overlap and correlation among the constructs. However, we ensure discriminant validity in measuring these constructs.

Another debate refers to the balance between completeness and parsimony of the input variables. Whereas McKelvey (1975, p. 514) recommends that researchers "define as many organizational attributes as possible," Mayr (1969) notes that there is little point in using more dimensions than are necessary to build a sound taxonomy. From a methodological angle, the presence of spurious dimensions (i.e., dimensions that do not differentiate among clusters) has been shown to have a detrimental effect on the performance of clustering methods. Punj and Stewart (1983, p. 143) caution "to avoid 'shotgun' approaches where everything known about the observations is used as the basis for clustering." Therefore, we distinguish between two types of variables in developing our taxonomy. First, we identify parsimonious sets of theory-based key constructs that serve as "active" input variables for the cluster algorithm. Second, we complement these with several "passive," nontheoretical, descriptive variables, which we use to characterize the clusters further.

Fundamental Dimensions of KAM

We begin our conceptualization of KAM by identifying the fundamental dimensions of the KAM phenomenon. Prior research on dimensions of KAM can be summarized in terms of three basic questions: ( 1) What is done? ( 2) Who does it? and ( 3) With whom is it done? However, as we have elaborated in the literature review, the scope of prior research has been limited to formalized key account programs with designated key account managers in place. We claim that to formalize or not to formalize the key account approach represents a decision dimension of its own. Therefore, we add a fourth question to KAM research: ( 4) How formalized is it? This leads us to conceptualize four dimensions of KAM. Drawing on research on the management of collaborative relationships that has distinguished among activities, actors, and resources (Anderson, Hákansson, and Johanson 1994; Narus and Anderson 1995), we refer to the four dimensions as ( 1) activities, ( 2) actors, ( 3) resources, and ( 4) formalization. The first dimension refers to interorganizational issues, and the other three refer to intraorganizational issues in KAM. Figure 1 visualizes our conceptualization of KAM.

Previous definitions of KAM have tended to focus on specific dimensions of KAM. Some authors focus on special activities for key accounts. For example, Barrett (1986, p.64) states that "National account management simply means targeting the largest and most important customers by providing them with special treatment in the areas of marketing, administration, and service." Others emphasize the dedication of special actors to key accounts. Yip and Madsen (1996,p. 24), for example, note that "National account management approaches include having one executive or team take overall responsibility for all aspects of a customer's business." Our conceptualization is more integrative because it encompasses activities and actors, as well as resources and formalization.

We now go through each of the four fundamental dimensions of KAM to identify parsimonious sets of theoretically based key constructs, which we use as (active) input variables for the cluster algorithm leading up to the taxonomy. We then identify additional descriptive (passive) variables that help enrich our descriptions of the clusters.

Activities. As we have shown, both the KAM literature (e.g., Lambe and Spekman 1997; Montgomery and Yip 2000; Napolitano 1997; Shapiro and Moriarty 1984b) and the relationship marketing literature suggest inventories of activities that suppliers can carry out for their key accounts. Among these are special pricing, customization of products, provision of special services, customization of services, joint coordination of the workflow, information sharing, and taking over business processes the customers outsources. The first question that arises with respect to organizational activities is how intensely they should be pursued. Shapiro and Moriarty (1980, p. 5) argue that "[a] key issue here is: How will or does the servicing of national accounts differ from that of other accounts?" Therefore, we define activity intensity as the extent to which the supplier does more for key accounts than for average accounts.

In addition to the level of intensity on an activity, another important conceptual issue is the origin of that intensity. Given that powerful customers often force their suppliers into special activities, the question arises whether the supplier or the key account proposes a special activity. Millman (1999, p. 2) observes that "some ... programs are seller-initiated, some are buyer-initiated." Empirical results by Sharma (1997) and Montgomery and Yip (2000) indicate that supplier firms indeed use KAM in response to customer demand for it. According to Arnold, Birkinshaw, and Toulan (1999, p. 15), "the proactive-reactive dimension matters a great deal." Therefore, we define activity proactiveness as the extent to which activities are initiated by the supplier.

Actors. Probably the most frequently discussed topic in key account program research is which special actors participate in key account activities. These specialized actors can be viewed as a personal coordination mode in KAM. The participation of special actors has a horizontal and a vertical component. The KAM literature suggests that there are many possibilities for horizontally placing KAM actors, ranging from a line manager who devotes part of his or her time to managing key accounts to teams that are fully dedicated to key accounts (Shapiro and Moriarty 1984a). Similarly, Olson, Walker, and Ruekert (1995) present a range of coordination mechanisms with a permanent team at one end of their continuum. Marshall, Moncrief, and Lassk (1999, p. 96) note that "team work is a fairly new concept in managing accounts and that salespeople are working in a team format much more today than in the past." Cespedes, Doyle, and Freedman (1989) even argue that selling is no longer an individual activity but rather a coordinated team effort. It has been suggested that the use of teams is a reaction to the use of purchasing teams on the buyer side (Hutt, Johnston, and Ronchetto 1985). We define the use of teams as the extent to which teams are formed to coordinate activities for key accounts.

Whereas teams refer to the horizontal participation in KAM, another fundamental issue pertains to vertical participation. The KAM actors may be placed at the headquarters, at the division level, or at the regional level (Shapiro and Moriarty 1984a). The importance of senior executive involvement in KAM has frequently been underscored in the KAM literature. As Millman and Wilson (1999, p. 330) note, KAM "is a strategic issue and the process should therefore be initiated and overseen by senior management." Napolitano (1997, p. 5) points out that "Top management must also play the lead role in securing business unit management support for the program." This view is supported by writers on strategy implementation, who argue that the organization is a reflection of its top managers (Hambrick and Mason 1984). Empirical support for the importance of top management has been provided by Jaworski and Kohli (1993), who find that market orientation is positively related to top management's emphasis on it. Therefore, we define top-management involvement as the extent to which senior management participates in KAM. The top-management involvement construct, adopted from the literature on strategy implementation and market orientation, is conceptually close to the centralization construct used in organization theory, which refers to the extent of decision authority that is concentrated on higher hierarchical levels.

Resources. As Shapiro and Moriarty (1984a, p. 2) note, "Much of the [national account management] concept as both a sales and a management technique revolves around the coordination of all elements involved in dealing with the customer." The KAM literature and the team-selling literature have pointed out that support is needed for key account activities from such diverse functional groups as marketing and sales, logistics, manufacturing, information technology, and finance and accounting (Moon and Armstrong 1994; Shapiro and Moriarty 1984b). "The key question, then, is: ... how can a salesperson obtain needed resources?" (Moon and Gupta 1997, p. 32). Obtaining resources has a pull and a push component.

In some cases, key account managers have special organizational power to ensure full cooperation from other organizational members. In other cases, key account managers must rely on their informal powers and interpersonal skills (Spekman and Johnston 1986, p. 522). Because the key account manager is typically part of the sales function (Shapiro and Moriarty 1984a), this lack of authority is most obvious for functional resources outside marketing and sales. We define access to nonmarketing and nonsales resources as the extent to which a key account manager can obtain needed contributions to KAM from nonmarketing and nonsales groups.

However, even within the marketing and sales function, a key account manager may face difficulty in receiving support for his or her tasks (Homburg, Workman, and Krohmer 1999; Platzer 1984). One common problem is the lack of authority over regional sales executives who handle the local business with global key accounts (Arnold, Birkinshaw, and Toulan 1999). For example, regional sales entities often resist companywide agreements on prices or service standards. Therefore, we define access to marketing and sales resources as the extent to which a key account manager can obtain needed contributions to KAM from marketing and sales groups.

Whereas access to resources refers to pulling on resources, research on team selling has frequently emphasized that the achievement of cross-functional integration in the selling center is facilitated if the participating functions themselves push cooperation (Smith and Barclay 1993). Day (2000, p. 24) notes that to develop strong relationships with customers, "a relationship orientation must pervade the mind-sets, values, and norms of the organization." Jaworski and Kohli (1993) refer to this concept of interdepartmental culture as esprit de corps. Culture is often viewed as a resource: "Organizational resources are the assets the firm possesses that arise from the organization itself, chief among these are the corporate culture and climate" (Morgan and Hunt 1999, p. 284). Fisher, Maltz, and Jaworski (1997) note that esprit de corps fosters the exchange of customer and market information. Therefore, we define the esprit de corps of the selling center as the extent to which selling center participants feel obliged to common goals and to each other.

Formalization. As Shapiro and Moriarty (1984a, p. 4) note, one of the "major organizational decisions that must be made as a company approaches a NAM program [is,] Should there be a NAM program or no program?" We believe that the distinction between more or less programmed approaches is highly relevant. As we show in our literature review, KAM approaches that do not have a key account program in place are underresearched.

Characteristics of KAM programs are the definition of reporting lines and formal linkages between departments, the establishment of formal expense budgets, the documentation of processes, and the development of formal guidelines for how to handle the accounts (Boles, Pilling, and Goodwyn 1994). In essence, the design decision of installing a key account program revolves around the extent to which KAM should be formalized. Consistent with writers on marketing organization (Olson, Walker, and Ruekert 1995; Workman, Homburg, and Gruner 1998), we define the formalization of a KAM approach as the extent to which the treatment of the most important customers is governed by formal rules and standard procedures. Thus, formalization can be viewed as an impersonal coordination mode, as opposed to top-management involvement and use of teams, which represent personal coordination modes in KAM.

Additional Descriptive Variables

In addition to the theoretical constructs developed previously, the KAM literature also suggests several descriptive variables to characterize KAM approaches. These variables refer to concrete, mostly demographic features of KAM approaches, such as the positions of key account managers. Because these variables are not theory based, we do not use them as input to the cluster procedure. However, given that these variables have frequently been discussed in KAM publications, we use them to enrich our interpretation of different KAM approaches subsequently.

In many companies, KAM teams are led by a key account manager. We define the key account coordinator as the person who is mainly responsible for coordinating activities related to key accounts. The first descriptive variable refers to the position of the key account coordinators. One possibility is to establish dedicated full-time positions for the coordination of key accounts (Pegram 1972). A fundamental question in this context is whether key account coordinators are placed in the supplier's headquarters or locally in the country or geographic region of the key account's headquarters. An alternative to the full-time option is a part-time responsibility. As Shapiro and Moriarty (1984a, p. 5) note, "the task is often taken on by top-level managers.... In other companies top marketing and sales managers and/or field sales managers take the responsibility." The second descriptive variable connects directly to this question of part-time versus full-time responsibility. We define the key account coordinators' dedication to key accounts as the percentage of their time they spend managing key accounts versus average accounts. Another question related to the allocation of time is how much time is spent with customers compared with the time devoted to internal coordination. Colletti and Tubridy (1987) report that 40% of a major account sales representative's time is administration work. We define the internal orientation of key account coordinators as the percentage of their time they spend on internal coordination versus external interaction with customers. A final descriptive question that has frequently been raised in KAM studies is how many accounts key account coordinators are typically looking after (Dishman and Nitse 1998; Sengupta, Krapfel, and Pusateri 1997a; Wotruba and Castle-berry 1993). We define the span of accounts as the number of accounts for which key account coordinators are responsible.

Outcomes

One of our objectives is to go beyond the conceptualization of KAM approaches and the taxonomy to explore the performance effects of design decisions. We distinguish between outcomes with respect to key accounts and outcomes on the level of the overall organization. Given that KAM involves investing in special activities and actors for key accounts that are not available for average accounts, we define KAM effectiveness as the extent to which an organization achieves better relationship outcomes for its key accounts than for its average accounts. Although the benefits of KAM have often been claimed in the KAM literature, empirical evidence on the outcomes of KAM is rare and methodologically limited to t-tests or correlations of single-item ratings of performance (Platzer 1984; Sengupta, Krapfel, and Pusateri 1997a; Stevenson 1981). A much better understanding of the outcomes of collaborative relationships has been developed by relationship marketing research (e.g., Kumar, Scheer, and Steenkamp 1995). This literature suggests that firms, through building relationships, pursue such outcomes as long-term orientation and continuity (e.g., Anderson and Weitz 1989; Ganesan 1994), commitment (e.g., Anderson and Weitz 1992; Geyskens et al. 1996; Gundlach, Achrol, and Mentzer 1995), trust (e.g., Geyskens, Steenkamp, and Kumar 1998; Moorman, Deshpandé, and Zaltman 1993; Rindfleisch 2000), and conflict reduction (e.g., Frazier, Gill, and Kale 1989).

Some authors indicate that KAM has outcomes not only with respect to key accounts but also at the organization level. As Cespedes (1993, p. 47) notes, "Another benefit is the impact on business planning. Salespeople at major accounts are often first in the organization to recognize emerging market problems and opportunities." Organization-level outcomes are also affected by average accounts. Following Ruekert, Walker, and Roering's (1985) terminology, we distinguish among adaptiveness, effectiveness, and efficiency. We define them as follows:

• Adaptiveness is the ability of the organization to change marketing activities to fit different market situations better than its competitors,
• Performance in the market is the extent to which the organization achieves better market outcomes than its competitors, and
• Profitability is the organization's average return on sales before taxes over the past three years.

Data Collection and Sample

Given our research objective of identifying prototypical approaches to KAM, we collected data using a mail survey in five business-to-business sectors in the United States and Germany. The questionnaire was initially designed in English and based on an extensive literature review and on field interviews with 25 managers, consultants, and academics in Germany and 25 in the United States on major trends in marketing organization (Homburg, Workman, and Jensen 2000). To ensure equivalent questionnaires in the two countries, the English version of the questionnaire was first translated into German by one expert translator and then retranslated into English by a second; both translators were bilingual. The two expert translators reconciled differences. We pretested the resulting two versions of the questionnaire and modified them in the United States and Germany on the basis of comments from 8 marketing and sales managers who completed the entire survey.

An important issue in designing our empirical study is obtaining the appropriate informants. We reiterate that the object of our research is the overall organizational approach toward the entire portfolio of key customers. A first implication of this is that, for the intraorganizational issues, the number of potential informants is limited to higher-level managers who have an overview over the marketing and sales organization. A second implication is that, regarding the outcomes of KAM, ideally the dyadic perceptions of all key accounts would need to be combined. In light of the obvious selection problems to obtain multiple, knowledgeable, high-level respondents as well as participation from several key accounts, we opted for a key informant approach. Although the single-respondent design curbs the generalizability of the results, John and Reve's (1982, p. 522) findings "indicate that careful selection of informants in conjunction with the use of internally consistent multi-item scales can provide reliable and valid data." On the basis of the field interviews, we determined that the most appropriate respondent is the head of the sales organization. We strove to minimize the limitation imposed by the single-informant design by determining the competence of the respondent to answer the survey. We excluded answers from lower-level respondents and from respondents with less than two years' experience in the selling organization from the analysis. As the description of our sample shows, our respondents are high-level managers.

We obtained a random sample of 1000 U.S. and 1000 German firms in the five business-to-business sectors from commercial list providers and sent an initial survey to the head of the sales organization. The cover letter and directions on the survey indicated that the survey should be answered by a vice president (VP) or director of sales or should be forwarded to someone familiar with how the firm's most important set of customers is managed. Because prior research has shown that managerial practice has different labels to denote important customers, we asked respondents to fill out the survey with respect to their most important set of business customers, regardless of the label they use for these customers. We sent a reminder postcard one week after the initial mailing to encourage response. We made follow-up telephone calls starting two weeks later to verify the contact name and the appropriateness of the firm for participation in the study and to encourage response. The survey was mailed a second time to all people approximately four weeks after the initial mailing. On the basis of the telephone calls and undeliverable mail, we determined that 174 of the U.S. firms and 171 of the German firms were inappropriate for the study. We received responses from 264 German firms and 121 U.S. firms, for effective response rates of 31.8% and 14.6%, respectively, and an overall response rate of 23.3% (for the sample composition, see Table 2). These response rates are in the range reported by other surveys sent to senior-level sales and marketing managers (Harzing 1997) and are comparable to the response rates of other data collections for taxonomic purposes (Bunn 1993; Cannon and Perreault 1999).

We controlled for a possible nonresponse bias in three ways. First, we divided the data into thirds in each country on the basis of the number of days from initial mailing to response (Armstrong and Overton 1977). The t-tests within each country between mean responses of early and late respondents indicated no statistically significant differences (p < .05). Second, we compared the German and the U.S. subsamples. The distributions in the subsamples do not differ statistically by revenue and by industry on the basis of chi-square tests (p > .05). Third, we compared the resulting KAM types to approaches identified in prior literature. As we elaborate in the "Results" section, we found that our taxonomy reflects all approaches to KAM that have been discussed previously. This supports the external validity of our taxonomy. We even detect several less formalized approaches that have not been described previously.

Measure Development Procedures

General measurement approach. Given the scarcity of prior empirical research, most scales for the study were newly generated. We used three types of measures in the survey: single-item measures, reflective multi-item measures, and formative multi-item measures. A single-item measure used in the survey was profitability. If observed variables (and their variances and covariances) were manifestations of underlying constructs, we used a reflective measurement model (Bagozzi and Baumgartner 1994). In that case, we can assess the scales' psychometric properties by means of criteria based on confirmatory factor analysis (Anderson and Gerbing 1988; Fornell and Larcker 1981). If necessary, we purified the item pools. Confirmatory factor analysis is considered superior to more traditional criteria (such as Cronbach's alpha) in the context of scale validation because of its less restrictive assumptions (AndersonandGerbing1988;Bagozzi, Yi, and Phillips 1991). We applied reflective measures if not otherwise indicated.

If a construct was a summary index of observed variables, a formative measurement model (Bagozzi and Baumgartner 1994) is more appropriate. In that case, observed variables cover different facets of the construct and cannot be expected to have significant intercorrelations. We used a formative scale to measure the proactiveness of activities for key accounts because, unlike intensity, the proactiveness on one activity item is not intercorrelated with the proactiveness on another. As an example, intense coordination of manufacturing schedules (high intensity) often requires highly coordinated logistics (high intensity). However, if a key account demands that the supplier coordinate manufacturing processes (low proactiveness), it may be the supplier who comes up with the suggestion to coordinate logistics as well in order to accomplish coordinated manufacturing (high proactiveness). Thus, although high intensity on one activity goes along with high intensity on another, this cannot be expected for proactiveness. The proactiveness construct must be understood in terms of a proactiveness index across the partial activities.

Control variables. In examining the performance effects of KAM, we have controlled for the effects of two environ-mental variables. Uncertainty has been identified as a determinant of performance in much of the research on organization theory and strategy. Specifically, we control for market dynamism. If customers' structures and needs change rapidly, it becomes more difficult for suppliers to be responsive to those needs. We also control for competitive intensity, which has been argued by many strategy researchers to be one of the most important determinants of performance (e.g., Porter 1980). Both control variables have frequently been employed in the related literature on market orientation (e.g., Jaworski and Kohli 1993; Pelham 1999).

Scale assessment. The Appendix provides our scale items and scale properties. We assessed measure reliability and validity using confirmatory factor analysis. Composite reliability represents the shared variance among a set of observed variables that measures an underlying construct (Fornell and Larcker 1981). Each construct manifests a composite reliability of at least .6 (Bagozzi and Yi 1988, p. 82). In addition, coefficient alpha values suggest a reasonable degree of internal consistency among the corresponding indicators. Nunnally (1978) recommends a threshold alpha value of .70 but suggests in a previous work (1967, p. 226) that a level of .6 is acceptable for exploratory research subjects (see also Murphy and Davidshofer 1988). For each of the KAM dimensions, outcomes, and control variables, we assessed discriminant validity on the basis of the criterion suggested by Fornell and Larcker (1981), which is recognized as more rigorous than the alternative chi-square difference test.

To ensure measurement invariance across countries, we followed the procedure suggested by Steenkamp and Baumgartner (1998). Given our objective to test dependence relationships among variables, configurational invariance and metric invariance must be fulfilled. Configurational invariance implies that the factorial structure underlying a set of observed measures is the same across the two countries. Metric invariance is a stricter criterion that assesses whether the units of measurement (i.e., the scale intervals) are equivalent in the German and the U.S. subsamples. Using multiple-group confirmatory factor analysis, we found full configurational invariance and at least partial metric invariance (at least two items were metric invariant) for our constructs. Therefore, merging the two national subsamples is valid.

Taxonomic Procedures

In the previous sections, we have identified fundamental dimensions of KAM approaches and have established rigorous measures of key constructs. Next, we give a brief summary of how we technically proceeded in identifying configurations of KAM on the basis of these key constructs. Given our objective of identifying prototypical approaches, we first decided to use nonoverlapping clustering and a distance measure. We followed the procedure used by Bunn (1993) and by Cannon and Perreault (1999) and took a multistage clustering approach. The two central issues in clustering are determining the appropriate number of clusters and assigning the observations to clusters.

We used the hierarchical clustering algorithm developed by Ward (1963) in combination with Sarle's (1983) cubic clustering criterion to determine the appropriate number of clusters. The cubic clustering criterion has been among the top-performing criteria in Milligan and Cooper's (1985) comparative study of 30 methods for estimating the number of population clusters. Ward's (1963) algorithm seeks at each step to form mutually heterogeneous and internally homogeneous clusters in the sense of the least error sum of squares. Because of the method's sensitivity to outliers, we standardized the clustering variables by dividing each variable by its range. Clustering ten randomly selected subsamples from our data, each containing two-thirds of the sample, we found strong support for an eight-cluster solution.[ 2] We also evaluated the stability of the result after eliminating outliers.

We then clustered the complete sample by means of a hybrid approach combining Ward's (1963) method with the k-means approach (Punj and Stewart 1983). Simulation studies on the performance of clustering algorithms demonstrate that partitioning methods (e.g., k-means) yield excellent results if given a reasonable starting solution (for an overview, see Milligan and Cooper 1987). Using Ward's method to compute a starting solution for k-means has been shown to be a powerful combination (Helsen and Green 1991) and has been recommended by Punj and Stewart (1983). Arabie and Hubert (1994, p. 169) note that "Nearly a decade later, that recommendation still seems like a good one." Finally, we cross-validated the stability of the cluster assignment using the procedure recommended by Cannon (1992).[ 3]

Taxonomy of Approaches to KAM

Given that we obtained the clusters on the basis of a purely technical procedure, we need to ensure that different clusters are not the consequence of different understandings of what an important account is. Therefore, we controlled for the importance of the criteria companies use to define and select their most important customers. For all clusters, the current and the potential sales volume dominates other criteria, such as learning about key technologies, the international scope of the account, the possibility of using the account as a reference, demand for special treatment by the account, or internal coordination problems in catering to the account. In conclusion, our statistical tests show that the clusters are comparable.

The last step in the taxonomy is to validate the recognizability of the clusters, which verifies whether they have meaningful interpretations (Rich 1992). Table 3 shows the cluster means for each of the eight cluster variables. Following the interpretation steps suggested by Bunn (1993), we first compared the clusters on the basis of Duncan's multiple-range test and then transferred the resulting bands into verbal descriptions of a cluster's position with respect to the cluster variables (see Table 4). The results for the additional descriptive variables are shown in Tables 3 and 5.

We now interpret the clusters in turn and assign labels to the approaches. Although there are risks of oversimplification in using such labels, they serve the didactic purpose of highlighting empirically distinct aspects of different approaches and facilitate the discussion of the results.

Top-management KAM. Top-management KAM truly deserves the name "program." These companies highly formalize the management of their key accounts. More than 60% of companies in this cluster have dedicated sales managers who coordinate activities for key accounts, which is consistent with the finding that 73% of key account coordinators' time is devoted to key accounts. Of the approaches, top-management KAM manifests the highest degree of top management involvement in KAM. Therefore, it is not surprising that this approach is managed out of the company headquarters (86.1% of key account coordinators are based in the suppliers' headquarters). In addition to heavy top management involvement, these companies make extensive use of teams. Activities for key accounts are intense and are proactively initiated. An interesting finding is that selling center esprit de corps is high, whereas access to marketing and sales as well as nonmarketing and nonsales resources is low. This may suggest that access to resources is barely needed. Top management might negotiate umbrella contracts, which operative teams carry out using highly standardized procedures.

Middle-management KAM. Middle-management KAM manifests a high level of formalization, but in contrast to the first approach, top-management involvement is medium. Intensity and proactiveness with respect to activities are also on a medium level. These results may suggest that these companies have installed a formal key account program, but on a middle-management level. Our interpretation is supported by the finding that 28.8% of key account coordinators are locally based in this approach, compared with 13.8% in top-management KAM. That key account managers are often locally based may also explain the high access to marketing and sales resources. On the contrary, selling center esprit de corps and access to nonmarketing and nonsales resources are low, which gives the overall impression that KAM in these companies is mainly driven by (local) middle management in the marketing and sales function.

Operating-level KAM. Companies using operating-level KAM are doing a lot for their key accounts and have considerably standardized procedures. In these aspects, this approach is comparable to top-management KAM and middle-management KAM. However, top-management involvement is lower than in these other approaches. Not surprisingly, access to functional resources is low. Whereas the VP of sales or marketing is the key account coordinator in 27.4% of top-management KAM companies and 23.1% of middle-management KAM companies, this is only the case for 9.8% of companies in the operating-level KAM cluster. The low degree of top-management involvement, along with fairly developed activities and teams, suggests that this KAM approach is mainly borne by the operating level. None of the other approaches has such a high percentage of companies with dedicated sales managers for key accounts (70.8%), 17.1% of whom are locally based.

Cross-functional, dominant KAM. The companies using cross-functional, dominant KAM have the highest values for nearly all variables. First, activities are intense and are proactively created. Second, formal procedures and team structures are fully developed. Top management is strongly involved. Third, selling center esprit de corps and access to functional resources are high. Of cross-functional KAM companies, 65.6% have dedicated sales managers as key account coordinators. Their share of time spent externally with the customer is the highest of all approaches, as is reflected by the 46% of time spent on internal orientation. The overall picture suggests that these companies are completely focused on their key accounts. It seems that, in these companies, customer management is virtually identical with KAM.

Unstructured KAM. As shown by the low values on formalization, top-management involvement, and use of teams, companies using unstructured KAM have not created special organizational structures for key accounts and do not have a program in place. This is consistent with the observation that activities are more a reaction than a proactive initiative, as is indicated by the 3.83 mean on proactiveness. Moreover, KAM comes mainly out of the headquarters, and key account coordinators are often normal sales managers (18.5% compared with 6.3% in cross-functional KAM). We observe that 62% of key account coordinator time is spent on internal coordination, the highest percentage of all clusters. This may account for the extremely high esprit de corps for KAM among selling center members and for the ease of obtaining contributions from marketing and sales as well as other functional resources. The overall impression is that these companies are pursuing KAM on an ad hoc basis, mobilizing internal resources only when the key accounts ask for it. Of these companies, 11.1% name the general manager to be the key account coordinator, though top management involvement is the lowest of all approaches. This suggests that the general management's responsibility exists on paper only.

Isolated KAM. Intensity and proactiveness of activities as well as formalization and use of teams manifest midrange values in the isolated KAM cluster. This implies that these companies are trying to do something for key accounts, which is supported by the finding that top management is fairly involved. The most striking feature is that in 44.4% of companies in this cluster, key account coordinators are locally based. This may explain why this cluster has low values on selling center esprit de corps and on access to nonmarketing and nonsales resources. Therefore, the overall picture is that KAM is a rather isolated, local sales effort in these companies that, despite some effort from the top management, struggles for cooperation from the central business units.

Country-club KAM. The striking characteristic of the country-club KAM cluster is a high degree of top management involvement that goes along with low values on most other variables. The management of key accounts in these companies is not guided by formal procedures, and teams are hardly ever formed. Special activities are performed less intensely and less proactively than under the other approaches. Most important, there are basically no dedicated key account coordinators. The KAM coordinator is often the VP of sales, a general manager, or even the VP of marketing. The comparatively low level of activities combined with high top-management involvement and high access to sales suggests that, in these companies, KAM is little more than representation by senior managers. In 33.3% of these firms, key accounts are simply handled by normal sales managers. With the exception of the top-management involvement, this approach is fairly close to the no-KAM cluster.

No KAM. The no-KAM cluster has the lowest values on nearly all variables: Comparatively little activity is performed, but not proactively. Formalization is low, as are cross-functional cooperation and esprit de corps. Mainly VPs of marketing and sales or general managers are named as key account coordinators, though top-management involvement in this cluster is low. This suggests that the VPs have responsibility on paper but do not actually perform that role. The interpretation of this approach is straightforward: These companies do not manage their key accounts. Or some companies may only have started to manage their key accounts, given that they profess to have dedicated key account coordinators.

Comparison with Existing Research

Although prior research has never classified KAM approaches empirically, there is some discussion of options companies have in implementing KAM. McDonald, Mill-man, and Rogers (1997) suggest ideal types of KAM, assuming that KAM approaches line up along a continuum from pre-KAM to synergistic KAM. Along the continuum, the activity intensity, the use of teams, and top-management involvement are assumed to rise, which implies a correlation among these design variables. Our results do not support this ideal continuum or the correlation. As we have shown, high degrees of top-management involvement occur in combination with both high and low degrees of activity intensity and in combination with both high and low degrees of use of teams.

Shapiro and Moriarty (1984a) propose another typology of KAM programs based on qualitative interviews in 19 large manufacturing and service companies (see also the supplementary comments by Kempeners and van der Hart [1999]). These researchers distinguish among six types of KAM programs that resemble the KAM approaches we identified. More specifically, their national account division resembles cross-functional KAM, their corporate-level program is similar to top-management KAM, their operating unit program at the group level is similar to middle-management KAM, their operating unit program at the division level parallels operating-level KAM, their part-time program resembles country-club KAM, and their noprogram option is close to the no-KAM approach. However, our work goes beyond the prior work by identifying the design variables behind the approaches, providing richer descriptions of the approaches, and supplementing the descriptions with quantitative data. We also detected two additional KAM approaches, unstructured KAM and isolated KAM. These two approaches involve a considerable number of activities for key accounts but do not require formalization of the approach. In conclusion, our findings seem to indicate that we have not overlooked KAM approaches that occur in practice. This speaks for the validity of our taxonomy and for the absence of a nonresponse bias.

Outcomes

We now turn to the success of the various KAM approaches. In interpreting the results in Table 6, we must pay attention to whether the outcome variable is on the level of the key accounts or of the organization as a whole. The effectiveness of KAM can be assumed to be strongly influenced by how key accounts are managed and is therefore our main out-come variable of interest. On the contrary, variance in organization-level outcomes, such as performance in the market, adaptiveness, and profitability, can be explained by many factors other than KAM. A firm may be driving its performance, for better or worse, through the average as opposed to the key accounts.[ 4]

On both the KAM level and the organization level, the no-KAM and the isolated KAM approaches perform the worst. On the organization-level outcomes, cross-functional KAM companies stand out with respect to both performance in the market and adaptiveness. As far as profitability is concerned, top-management KAM companies perform best. That the most effective approaches are not the most profitable ones may be explained by some approaches involving higher costs in addition to generating higher revenues.

Another observation in Table 6 is that several KAM approaches are equally successful. This finding is consistent with the concept of "equifinality" emphasized in the configurational approach (Meyer, Tsui, and Hinings 1993). However, given our key informant design, it raises the issue whether a common method bias is present in the data. Two facts from our data speak against the presence of a bias. First, a possible key informant bias should affect the subjective performance measures (e.g., KAM effectiveness), but not the objective performance measure (i.e., profitability). That several configurations also manifest the same level of objective performance supports the validity of our findings on the subjective measures. Second, even in very active approaches (e.g., top-management KAM), there is much variance across the respondents regarding the performance variables. Indeed, the lack of significant differences among some approaches is due to the high variance rather than a tendency of all key informants to rate their own approach highly.

It is necessary to verify whether the performance differences hold true even when we consider environmental variables. Market dynamism and competitive intensity have been shown to influence performance in a market orientation context (Jaworski and Kohli 1993). To control for these effects, we made use of analysis of covariance (ANCOVA). Cluster membership was the (nominal) factor, and the control variables served as covariates. Table 7 shows that though market dynamism has a significant effect on performance in the market and competitive intensity has an effect on profitability, the effects of cluster membership on all performance outcomes are still significant.

Research Contribution

Despite the immense importance of KAM in managerial practice, prior research in this area has been fragmented, and sound empirical studies have been scarce. The contributions of this article come from both the conceptualization and the taxonomy.

The first contribution of this article is to provide conceptual clarity to KAM design decisions and to lay the basis for further research. In addition to synthesizing the existing literature, this article extends the conceptual scope of KAM research by drawing attention to the failure of previous research to go beyond the boundaries of formalized KAM programs and study nonformalized KAM approaches. We derive an integrative conceptualization of KAM that identifies four key dimensions: ( 1) activities, ( 2) actors, ( 3) resources, and ( 4) formalization (see Figure 1). We also develop scales for key constructs related to KAM.

A second contribution of our work consists in its being the first study to empirically classify designs of organizational approaches to selling. Although taxonomies exist for the buyer side (Bunn 1993) and for the relationship between buyer and seller (Cannon and Perreault 1999), there has been no taxonomy on the organization of the seller side. Moncrief (1986) has created a taxonomy of individual sales position designs, but the level of analysis in selling research has shifted to the selling team (Weitz and Bradford 1999). As Marshall, Moncrief, and Lassk (1999, p. 88) state, "Clearly, the operative set of sales activities representing a sales job in the mid-1980s is deficient to accurately under-stand and portray sales jobs of today." Therefore, our taxonomy closes a gap in empirical knowledge about organizational approaches to selling.

A third major contribution is the refinement of existing KAM typologies. We confirmed the types of KAM postulated by Shapiro and Moriarty (1984a), supplemented them with empirical detail, and detected two additional approaches. These two involve a considerable number of activities for key accounts but do not require formalization of the approach.

An additional contribution of our taxonomic research is to provide deeper insights into the performance aspects of KAM approaches. On a general level, it is important to note that the same level of performance can be accomplished through different approaches. Yet some approaches perform significantly worse than others. The finding that no-KAM companies are behind on all performance dimensions represents the most comprehensive empirical demonstration so far that suppliers benefit from managing their key accounts. The similar performance of isolated KAM indicates that mediocre approaches to KAM are likely to fail. These results suggest that failure to achieve access to and commitment of cross-functional resources seems to play a critical role for the success of KAM programs. This reinforces recent research on marketing organization that recognizes the cross-functional dispersion of marketing activities (Workman, Homburg, and Gruner 1998).

On a general level, our work has shown that there is value in blending relationship marketing concepts and marketing organization concepts. Within our conceptual model, the actor, resources, and formalization dimensions are inspired by marketing organization research, and the activity and the outcome dimensions draw on relationship marketing research.

Avenues for Further Research

Further research should continue building the bridge between relationship marketing concepts and marketing organization concepts. One possible avenue is to empirically link the KAM approaches identified in this article to relationship types (Cannon and Perreault 1999). In designing these empirical studies, the existence of nonformalized KAM approaches should be carefully considered.

Future empirical designs should also seek to overcome some of the limitations of this article. One limitation stems from the static design of our study. As research by Pardo, Salle, and Spencer (1995) has shown, key account approaches evolve over time. Further research should also capture the dynamic performance effects of KAM. As Kalwani and Narayandas (1995) have shown, the beneficial outcomes of customer-oriented activities appear with a certain delay. Another limitation of our article is the use of a single-informant design, which focuses on one side of the seller-buyer dyad. Future studies should also take the key accounts' perspectives into consideration. This is particularly important for analyzing the outcomes of KAM. One way to extend our examination of outcomes would be to differentiate the performance impacts of individual KAM dimensions. In this context, the effect of KAM-level outcomes on organization-level outcomes should be explored as well.

Another open issue is the effect of the environment on KAM dimensions. The literature has claimed that the formation of key account programs is influenced by characteristics of buyers and of the market environment, such as purchasing centralization, purchasing complexity, demand concentration, and competitive intensity (Boles, Johnston, and Gardner 1999; Stevenson 1980). Yet rigorous empirical research linking multiple environmental dimensions to multiple KAM dimensions is still lacking.

Managerial Implications

One of the most fundamental managerial tasks is designing the internal organization. These design decisions are typically taken on the level of the organization rather than the level of individual accounts. Therefore, the organizational perspective adopted in this research has particular appeal to top executives.

The key message to managers is not to take a laissez-faire approach to KAM. Given that the no-KAM option is markedly less successful than other approaches, our results call for managers to manage key accounts actively. That there are significant performance differences among the more actively managed approaches demonstrates that it is important to design the approach in detail. Our work also shows that KAM requires support from the whole organization. Therefore, top managers should not leave the design of the KAM approach to the sales organization alone.

The conceptualization of KAM developed in this article provides managers with a systematic way to design the KAM approach. As Day and Montgomery (1999, p. 12) note, "conceptual frameworks, typologies, and metaphors that are the precursors to actual theory building" provide valuable guidelines for managers. Managers should work through four questions: ( 1) What should be done for key accounts? ( 2) Who should do it? ( 3) With whom in the organization is cooperation needed? and ( 4) How formalized should the KAM approach be? We particularly emphasize that managing key accounts does not necessarily require setting up a formal key account program.

The taxonomy developed in this article further supports managers in designing their KAM. Managers can categorize their own companies' approach on the basis of the prototypical implementation forms identified. From the taxonomy, they can discover neglected design areas and develop alternative designs.

Key account management is a highly relevant issue for marketing and sales managers. In addition, it is an area for academic research, because it builds a bridge between marketing organization and relationship marketing. Therefore, the lack of sound academic research in this area is surprising. This article provides the basis for further research by contributing an integrative conceptualization of KAM. It also fills a gap in knowledge about how firms design their approach to key accounts. Finally, it shows that actively managing key accounts leads to significantly better performance than neglecting them does.

1 It is worth noting that some companies use different labels to denote various degrees of an account's importance within a key account program (Napolitano 1997; Shapiro and Moriarty 1982).

• 2 Seven subsamples manifested eight clusters, one manifested seven clusters, and two manifested no-cluster structure according to the cubic clustering criterion for a range of one to ten clusters.
• 3 We split the sample into three equally large subsamples (A, B, and C) and ran through the hybrid approach twice for [Multiple line equation(s) cannot be represented in ASCII text; please see PDF of this article if available.] and [Multiple line equation(s) cannot be represented in ASCII text; please see PDF of this article if available.] We then evaluated whether observations in Subsample B had been assigned to the same cluster in both runs.
• 4 We owe this idea to an anonymous reviewer.

Legend for chart:
A = Authors
B = Year
C = Empirical Basis
D = Dimensions Discussed
E = Main Focus/Key Statements


A Boles, Barksdale, and Johnson
B 1996
C 73 national account decision makers from NAMA list
D
E • Identifies salesperson activities, skills, and attitudes
  that are appreciated by key account decision makers.

A Weeks and Stevens
B 1997
C 133 NAMA members
D
E • Key account managers are dissatisfied with sales training
  programs.
  • Descriptives on experience and skills of key account
  managers.


Group 2 Research on Key Account Relationships

A Lambe and Spekman
B 1997
C 118 managers, mostly U.S. based
D
E • Explores differences between national account relationships
  and other types of strategic alliances.

A McDonald, Millman, and Rogers
B 1997
C Interviews with 11 key account manager/purchasing manager dyads
D
E • Describes development of key account relationships from
  pre-KAM, transactional phase to collaborative relationship that
  goes along with increasing complexity of involvement.

A Pardo
B 1997
C 20 interviews with key accounts of electricity and telephone
  companies
D
E • Suggests three ways that key accounts perceive KAM;
  disenchantment, interest, and enthusiasm.
  • Moderators of KAM program perception by customers are
  perceived product importance and centralization of purchase
  decisions.

A Sengupta, Krapfel, and Pusateri
B 1997b
C 176 NAMA members in manufacturing and service companies
D
E • Switching costs in key account relationships.

A Sharma
B 1997
C 109 interviews with buyers of telephone equipment
D
E • Customers' preference for KAM programs depends on levels
  involved in purchasing, functions involved in purchasing, and time
  taken for purchasing.


Group 3: Research on KAM Approaches

A Colletti and Tubridy
B 1987
C 105 NAMA members
D Actors
E • Explores reporting level, time utilization, compensation,
  and required skills of account managers.

A Dishman and Nitse
B 1998
C 27 interviews with NAMA members whose key account program is older
  than five years
D Actors
E • Implementation options of national account management are
  cooperation with existing sales force, company executives, or a
  separate sales force.
  • Descriptives on number and size of customers in KAM
  program.

A Montgomery and Yip
B 2000
C 191 managers in 165 manufacturing and service companies
D Activities, actors, outcomes of KAM
E • Use of global account management structures will increase.
  • Use of global account management structures is driven by
  customer demand.
  • Customer demands encompass coordination of resources,
  uniform terms of trade, and consistency in service quality and
  performance.

A Napolitano
B 1997
C NAMA study among Fortune- 1000 companies, no sample size provided
D Actors, outcomes of KAM
E • The number of national account managers has tripled between
  1992 and 1996.
  • 53% of companies report poor effectiveness of partnering
  with customers.

A Pardo, Salle, and Spencer
B 1995
C 10 interviews within one telecom company
D Activities, actors, resources
E • Case study of one key account program over 20 years.

A Pegram
B 1972
C 250 interviews with executives in manufacturing and service
  companies
D Activities, actors
E Describes alternatives for assigning KAM responsibility on a
  part-time or a full-time basis.

A Platzer
B 1984
C 130 interviews with national account executives
D Activities, actors, resources, outcomes of KAM
E • Describes activities for key accounts.
  • Describes types of national account units.
  • Describes success factors of national account programs.

A Sengupta, Krapfel, and Pusateri
B 1997a
C 176 NAMA members in manufacturing and service companies
D Actors, outcomes of KAM
E • Descriptive statistics on growth of KAM approaches and key
  account manager workload.
  • Identifies customer-based compensation as a success factor
  of KAM.

A Shapiro and Moriarty
B 1984a
C 100+ interviews in 19 manufacturing and service companies
D Actors
E • Describes alternatives for integrating a KAM program into
  the structural organization.
  • Discusses issues pertaining to the internal structure of
  KAM units.

A Shapiro and Moriarty
B 1984b
C 100+ interviews in 19 manufacturing and service companies
D Activities, resources
E • Describes customer need for activities in such areas as
  pricing, products, service, and information.
  • Describes roles of various functional groups in the
  performance of activities for key accounts.

A Stevenson
B 1981
C 34 executives in 33 manufacturing companies
D Actor, outcomes of KAM
E • Explores payoffs from national account management.

A Wotruba and Castleberry
B 1993
C 107 NAMA members
D Actors, outcomes of KAM
E • Explores staffing procedures for KAM positions.
  • Performance of key account managers is affected by length
  of tenure, age of program, and time devoted to key accounts.

A Yip and Madsen
B 1996
C Case studies of IBM, AT&T, and Hewlett-Packard
D Actors, resources
E • Develops framework for global account management.
  • Describes internal cooperation for key accounts in global
  companies.

Notes: NAMA = National Account Management Association.

                                                              Total
A: Position of Respondents                                   (n=385)

Managing director, CEO, VP of region, head of business unit    19%
VP marketing, VP sales, VP sales and marketing                 49%
Head of KAM, key account manager                                9%
Sales manager, product manager                                 19%
Other                                                           3%
                                                      United
                                           Germany  States    Total
B: Demographics of the Firms               (n=264)  (n=121)  (n=385)

Industry*      Chemical and pharmaceutical   24%      18%      22%
                        Machinery            22%      30%      25%
                Computer and electronics     17%      14%      16%
                  Banks and insurances       17%      11%      15%
                Food and packaged goods      20%      27%      22%

Annual Revenues*    <$15 million           5%      10%       6%
                    $15-$30 million          14%      11%      13%
                    $30-$60 million          20%      15%      18%
                    $60-$150 million         17%      24%      19%
                   $150-$300 million         13%      11%      13%
                   $300-$600 million         11%      13%      12%
                 $600-$1,500 million          5%      10%       6%
                  >$1,500 million         14%      11%      13%

*Equal structure of subsamples based on p(χ²) > .05.

Legend for chart:
A = Dimension
B = Variable
C = Cluster: Top Management KAM (n=37)
D = Cluster: Middle Management KAM (n=76)
E = Cluster: Operating Level KAM (n=57)
F = Cluster: Cross functional, dominant KAM (n=44)
G = Cluster: Unstructured KAM (n=38)
H = Cluster: Isolated KAM (n=40)
I = Cluster: Country-Club KAM (n=37)
J = Cluster: No KAM (n=46)
K = Cluster: Total (n=375)

    A            B              C           D          E
                                F           G          H
                                I           J          K

Activities   Activity        5.08[bc]    4.99[b]    5.15[bc]
              intensity      5.44[c]     4.75[b]    5.00[b]
                             4.19[a]     4.11[a]    4.86

             Activity        4.15[bc]    4.13[bc]   4.27[ab]
              proactiveness  4.60[d]     3.83[ab]   4.16[bc]
                             3.54[a]     3.79[ab]   4.08

Formali-     Approach        5.48[f]     5.05[e]    4.58[d]
zation        formalization  5.64[f]     2.81[b]    3.64[c]
                             2.12[a]     2.72[b]    4.15

Actors       Top management  5.66[e]     3.98[c]    3.19[b]
              involvement    4.48[d]     2.52[a]    4.23[cd]
                             4.59[d]     3.19[b]    3.93

             Use of teams    5.05[d]     3.08[b]    5.32[de]
                             5.62[e]     3.16[b]    4.49[c]
                             2.18[a]     2.53[a]    3.93

Resources    Selling         5.57[c]     5.28[b]    5.52[c]
              center esprit  6.14[d]     5.97[d]    3.93[a]
              de corps       4.69[b]     3.82[a]    5.14

             Access to       5.34[bc]    5.82[de]   5.11[b]
              marketing and  6.51[f]     5.95[d]    5.50[cd]
              sales          6.44[f]     4.48[a]    5.62
              resources

             Access to       4.37[a]     5.18[b]    4.42[a]
              nonmarketing   6.05[c]     5.51[b]    4.29[a]
              and nonsales   5.40[b]     4.13[a]    4.92
              resources

Additional   Dedication      73%[c]      66%[abc]   70%[bc]
Descriptive   to key         73%[c]      57%[a]     66%[abc]
Variables     accounts       62%[abc]    57%[ab]    66%

             Internal        50%[ab]     49%[ab]    49%[ab]
              orientation    46%[a]      62%[c]     51%[ab]
                             49%[ab]     58%[bc]    51%

             Span of         5           5          5
              accounts       5           5          5
              (median)       8           10         5

Notes: Reported values are mean values if not otherwise noted. In each row, cluster means that have the same superscript are not significantly different (p < .05) on the basis of Duncan's multiple-range test. Means in the lowest band are assigned "a," means in the next highest band "b," and so forth. Means in the highest band are printed in bold; means in the lowest band are in italics.

Legend for chart:
A = Variable
B = KAM Approach: Top-Management KAM (n=37)
C = KAM Approach: Middle-Management KAM (n=76)
D = KAM Approach: Operating-Level KAM (n=57)
E = KAM Approach: Cross-functional, dominant KAM (n=44)
F = KAM Approach: Unstructured KAM (n=38)
G = KAM Approach: Isolated KAM (n=40)
H = KAM Approach: Country-Club KAM (n=37)
I = KAM Approach: No KAM (n=46)
J = KAM Approach: Total (n=375)

 A       B       C       D      E        F      G      H      I     J

Acti-  *Med-    Med-  *Medium *High    Medium  Med-  ^Low   ^Low
vity    ium     ium    high                    ium
inten-  high
sity

Acti-   Medium  Med-  ^Low-   *High   ^Low-    Med-  ^Low   ^Low
vity            ium    medium          medium  ium           medium
pro-
active-
ness

App-   *Very    High   Rather *Very    Low     Rath- ^Very   Low
roach   high           high    high            er     low
forma-                                         low
liza-
tion

Top-   *Very    Med-   Low     High   ^Very    Med-   High   Low
man-    high    ium                    low     ium
age                                            high
ment
in-
volve-
ment

Use of  Much    Lit-  *Much-  *Very    Little  Med-  ^Very  ^Very
teams           tle    very    much            ium    lit-   little
                       much                           tle

Sell-   Rather  Rath-  Rather *Strong *Strong ^Weak   Rath- ^Weak
ing     strong  er     strong                         er
center          weak                                  weak
esprit
de
corps

Access  Rather  High   Low    *Very    Rather  Med-  *Very  ^Very
to      low                    high    high    ium    high   low
mar-
keting
and
sales
re-
sources

Access ^Low     Med-  ^Low    *High    Medium ^Low    Med-  ^Low
to non-         ium                                   ium
market-
ing
and
non-
sales
re-
sources

Notes: Means in the highest band have *; means in the lowest band have ^.

Legend for chart:
A = Position of Key Account Coordinator
B = KAM Approach: Top-Management KAM (n=37) Headquarter
C = KAM Approach: Top-Management KAM (n=37) Local
D = KAM Approach: Middle-Management KAM (n=76) Headquarter
E = KAM Approach: Middle-Management KAM (n=76) Local
F = KAM Approach: Operating Level KAM (n=57) Headquarter
G = KAM Approach: Operating Level KAM (n=57) Local
H = KAM Approach: Cross-functional, dominant KAM (n=44) Headquarter
I = KAM Approach: Cross-functional, dominant KAM (n=44) Local
J = KAM Approach: Unstructured KAM (n=38) Headquarter
K = KAM Approach: Unstructured KAM (n=38) Local
L = KAM Approach: Isolated KAM (n=40) Headquarter
M = KAM Approach: Isolated KAM (n=40) Local
N = KAM Approach: Country-Club KAM (n=37) Headquarter
O = KAM Approach: Country-Club KAM (n=37) Local
P = KAM Approach: No KAM (n=46) Headquarter
Q = KAM Approach: No KAM (n=46) Local
R = KAM Approach: Total (n=375)

  A
  B        C       D       E       F       G      H      I       J
  K        L       M       N       O       P      Q      R

Normal sales manager
 3.4%     --     5.8%    3.8%    2.4%    7.3%    6.3%   --     14.8%
 3.7%    7.4%   11.1%   19.0%   14.3%    8.3%    8.3%   14%

Dedicated sales manager
48.3%   13.8%   44.2%   17.3%   53.7%   17.1%   40.6%   25.0%  44.4%
 7.4%   18.5%   33.3%    --      4.8%   22.2%    5.6%   52%

VP of sales
24.1%    --     17.3%    5.8%    9.8%    --     15.6%   --     11.1%
 3.7%   14.8%    --     28.6%    9.5%   30.6%    5.6%   22%

VP of marketing
 3.4%    --      --      --       --     --      3.1%   --      3.7%
  --     3.7%    --      9.5%     --     8.3%     --    37%

General manager
  --      --     1.9%    1.9%    4.9%    --       --    --     11.1%
  --     3.7%    --      9.5%     --     8.3%     --    57%

Other
 6.9%     --     1.9%     --     4.9%    --      9.4%   --       --
  --     7.4%    --      4.8%     --     2.8%    --     57%

Total
86.1%   13.8%   71.1%   28.8%   75.7%   24.4%   75.0%   25.0%  85.1%
14.8%   55.5%   44.4%   71.4%   28.6%   80.5%   19.5%

Legend for chart:
A = Level
B = Variable
C = KAM Approach Top-Management KAM (n=37)
D = KAM Approach Middle-Management KAM (n=76)
E = KAM Approach Operating Level KAM (n=57)
F = KAM Approach Cross-functional, dominant KAM (n=44)
G = KAM Approach Unstructured KAM (n=38)
H = KAM Approach Isolated KAM (n=40)
I = KAM Approach Country-Club KAM (n=37)
J = KAM Approach No KAM (n=46)
K = KAM Approach Total (n=375)

   A            B            C           D           E
                             F           G           H
                             I           J           K

KAM       KAM             5.39[b]*     5.39[b]*    5.53[b]*
          effectiveness   5.63[b]*     5.46[b]*    5.01[a]^
                          5.41[b]*     5.04[a]^    5.37

Overall   Performance     5.03[bc]     5.23[cd]*   5.04[bc]
organi-   in the market   5.51[d]*     5.19[cd]*   4.72[ab]^
zation                    5.16[cd]*    4.54[a]^    5.07

          Adaptiveness    4.75[bc]     4.87[b]     4.46[ab]^
                          5.43[d]*     4.85[bc]    4.25[a]^
                          4.50[abc]^   4.23[a]^    4.68

          Profitability   6.38[b]*     4.98[a]^    4.98[a]^
                          5.64[ab]*    5.84[ab]*   5.23[ab]*
                          4.82[a]^     4.80[a]^    5.27

Notes: Reported values are mean values if not otherwise noted. In each row, cluster means that have the same [superscript] are not significantly different (p < .05) on the basis of Duncan's multiple-range test. Means in the lowest band are assigned "a," means in the next highest band "b," and so forth. Means in the highest band have *; means in the lowest band have ^.

Legend for chart:
A = Dependent Variable: Models 1 Through 4
B = ANCOVA Results Total Model Mean of Squares
C = ANCOVA Results Total Model F (p)
D = ANCOVA Results Approach Mean of Squares (d.f. = 7)
E = ANCOVA Results Approach F (p)
F = ANCOVA Results Covariates Market Dynamism Mean of Squares
    (D.f. = 1)
G = ANCOVA Results Covariates Market Dynamism F (p)
H = ANCOCA Results Covariates Competitive Intensity Mean of Squares
    (d.f.=1)
I = ANCOVA Results Covariates Competitive Intensity F (p)

      A           B        C      D      E      F    G       H    I

1. KAM          1.76     3.72   2.20   4.65  <.01  <.01    .81  1.72
   effect- (d.f.=9;370) (<.01)        (<.01)       (.96)        (.19)
   iveness
2. Perform-     3.63     5.58   4.15   6.37  2.27  3.48    .85  1.31
   ance in (d.f.=9;370) (<.01)        (<.01)       (.06)  (.25) (.25)
   the
   market
3. Adapt-       5.72     7.81   6.54   8.93  1.93  2.64    .47   .64
   iveness (d.f.=9;369) (<.01)        (<.01)       (.11)        (.42)
4. Profit-     22.81     3.99  12.19   2.13   .90   .16  125.97 22.05
   ability (d.f.=9;322) (<.01)         (.04)       (.69)       (<.01)

Notes: d.f. = degrees of freedom.

DIAGRAM: FIGURE 1 Conceptualization of KAM

Anderson, Erin and Barton Weitz (1989), "Determinants of Continuity in Conventional Industrial Channel Dyads," Marketing Science, 8 (4), 310-23.

--- and --- (1992), "The Use of Pledges to Build and Sustain Commitment in Distribution Channels," Journal of Marketing Research, 29 (February), 18-34.

Anderson, James C. and David W. Gerbing (1988), "Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach," Psychological Bulletin, 103 (May), 411-23.

---, Hákan Hákansson, and Jan Johanson (1994), "Dyadic Business Relationships Within a Business Network Context," Journal of Marketing, 58 (October), 1-15.

Arabie, Phipps and Lawrence Hubert (1994), "Cluster Analysis in Marketing Research," in Advanced Methods of Marketing Research, Richard P. Bagozzi, ed. Cambridge, MA: Blackwell, 160-79.

Armstrong, J. Scott and Terry S. Overton (1977), "Estimating Non-response Bias in Mail Surveys," Journal of Marketing Research, 14 (August), 396-402.

Arnold, David, Julian Birkinshaw, and Omar Toulan (1999), "Implementing Global Account Management in Multinational Corporations," Thexis, 16 (4), 14-17.

Bagozzi, Richard P. and Hans Baumgartner (1994), "The Evaluation of Structural Equation Models and Hypothesis Testing," in Principles of Marketing Research, Richard P. Bagozzi, ed. Cambridge, MA: Blackwell, 386-422.

--- and Youjae Yi (1988), "On the Evaluation of Structural Equation Models," Journal of the Academy of Marketing Science, 16 (1), 74-94.

---, ---, and Lynn W. Phillips (1991), "Assessing Construct Validity in Organizational Research," Administrative Science Quarterly, 36 (September), 421-58.

Bailey, Kenneth D. (1994), Typologies and Taxonomies: An Introduction to Classification Techniques. Thousand Oaks, CA: Sage Publications.

Barrett, John (1986), "Why Major Account Selling Works," Industrial Marketing Management, 15 (1), 63-73.

Boles, James S., Hiram C. Barksdale Jr., and Julie T. Johnson (1996), "What National Account Decision Makers Would Tell Salespeople About Building Relationships," Journal of Business & Industrial Marketing, 11 (2), 6-19.

---, Wesley Johnston, and Alston Gardner (1999), "The Selection and Organization of National Accounts: A North-American Perspective," Journal of Business & Industrial Marketing, 14 (4), 264-75.

---, Bruce K. Pilling, and George W. Goodwyn (1994), "Revitalizing Your National Account Marketing Program: The NAM-Audit," Journal of Business & Industrial Marketing, 9 (1), 24-33.

Bunn, Michele D. (1993), "Taxonomy of Buying Decision Approaches," Journal of Marketing, 57 (January), 38-56.

Cannon, Joseph P. (1992), "A Taxonomy of Buyer-Seller Relationships in Business Markets," doctoral dissertation, Kenan-Flagler Business School, University of North Carolina-Chapel Hill.

--- and William D. Perreault Jr. (1999), "Buyer-Seller Relationships in Business Markets," Journal of Marketing Research, 36 (November), 439-60.

Cardozo, Richard N., Shannon H. Shipp, and Kenneth J. Roering (1987), "Implementing New Business-to-Business Selling Methods," Journal of Personal Selling & Sales Management, 7 (August), 17-26.

---, ---, and --- (1992), "Proactive Strategic Partner-ships: A New Business Markets Strategy," Journal of Business & Industrial Marketing, 7 (1), 51-63.

Cespedes, Frank V. (1993), "Coordinating Sales and Marketing in Consumer Goods Firms," Journal of Consumer Marketing, 10 (2), 37-55.

---, Stephen X. Doyle, and Robert J. Freedman (1989), "Teamwork for Today's Selling," Harvard Business Review, 67 (March/April), 44-54.

Colletti, Jerome A. and Gary S. Tubridy (1987), "Effective Major Account Sales Management," Journal of Personal Selling and Sales Management, 7 (August), 1-10.

Day, George S. (2000), "Managing Market Relationships," Journal of the Academy of Marketing Science, 28 (Winter), 24-30.

---and David B. Montgomery (1999), "Charting New Directions for Marketing," Journal of Marketing, 63(Special Issue),3-13.

Dishman, Paul and Philip S. Nitse (1998), "National Accounts Revisited: New Lessons from Recent Investigations," Industrial Marketing Management, 27 (January), 1-9.

Dorsch, Michael J., Scott R. Swanson, and Scott W. Kelley (1998), "The Role of Relationship Quality in the Stratification of Vendors as Perceived by Customers," Journal of the Academy of Marketing Science, 26 (2), 128-42.

Fisher, Robert J., Elliot Maltz, and Bernard J. Jaworski (1997), "Enhancing Communication Between Marketing and Engineering: The Moderating Role of Relative Functional Identification," Journal of Marketing, 61 (July), 54-70.

Fornell, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error," Journal of Marketing Research, 18 (February), 39-50.

Frazier, Gary L., James D. Gill, and Sudhir H. Kale (1989), "Dealer Dependence Levels and Reciprocal Actions in a Channel of Distribution in a Developing Country," Journal of Marketing, 53 (January), 50-69.

Ganesan, Shankar (1994), "Determinants of Long-Term Orientation in Buyer-Seller Relationships," Journal of Marketing, 58 (April), 1-19.

Geyskens, Inge, Jan-Benedict E.M. Steenkamp, and Nirmalya Kumar (1998), "Generalizations About Trust in Marketing Channel Relationships Using Meta-Analysis," International Journal of Research in Marketing, 15 (July), 223-48.

---, ---, Lisa K. Scheer, and Nirmalya Kumar (1996), "The Effects of Trust and Interdependence on Relationship Commitment: A Trans-Atlantic Study," International Journal of Research in Marketing, 13 (4), 303-17.

Gundlach, Gregory T., Ravi S. Achrol, and John T. Mentzer (1995), "The Structure of Commitment in Exchange," Journal of Marketing, 59 (January), 78-92.

Hambrick, Donald C. and Phyllis A. Mason (1984), "Upper Echelons: The Organization as a Reflection of Its Top Managers," Academy of Management Review, 9 (2), 193-206.

Harzing, Anne-Will (1997), "Response Rates in International Mail Surveys: Results of a 22-Country Study," International Business Review, 6 (6), 641-65.

Helsen, Kristiaan and Paul E. Green (1991), "A Computational Study of Replicated Clustering with an Application to Market Segmentation," Decision Sciences, 22 (November/December), 1124-41.

Homburg, Christian, John P. Workman Jr., and Ove Jensen (2000), "Fundamental Changes in Marketing Organization: The Movement Toward a Customer-Focused Organizational Structure," Journal of the Academy of Marketing Science, 28(Fall),459-78.

---, ---, and Harley Krohmer (1999), "Marketing's Influence Within the Firm," Journal of Marketing, 63 (April), 1-17.

Hunt, Shelby D. (1991), Modern Marketing Theory. Cincinnati, OH: South-Western Publishing.

Hutt, Michael D., Wesley J. Johnston, and John R. Ronchetto Jr.

(1985), "Selling Centers and Buying Centers: Formulating Strategic Exchange Patterns," Journal of Personal Selling & Sales Management, 5 (May), 33-40.

Jaworski, Bernard J. and Ajay K. Kohli (1993), "Market Orientation: Antecedents and Consequences," Journal of Marketing, 57 (July), 53-70.

John, George and Torger Reve (1982), "The Reliability and Validity of Key Informant Data from Dyadic Relationships in Marketing Channels," Journal of Marketing Research, 19 (November), 517-24.

Jolson, Marvin A. (1997), "Broadening the Scope of Relationship Selling," Journal of Personal Selling & Sales Management, 17 (4), 75-88.

Kalwani, Manohar and Narakesari Narayandas (1995), "Long-Term Manufacturer-Supplier Relationships," Journal of Marketing, 59 (January), 1-16.

Kempeners, Marion A. and Hein W. van der Hart (1999), "Designing Account Management Organizations," Journal of Business & Industrial Marketing, 14 (4), 310-27.

Kumar, Nirmalya, Lisa K. Scheer, and Jan-Benedict E.M. Steenkamp (1995), "The Effects of Perceived Interdependence on Dealer Attitudes," Journal of Marketing Research, 32 (August), 348-56.

Lambe, Jay C. and Robert E. Spekman (1997), "National Account Management: Large-Account Selling or Buyer-Supplier Alliance?" Journal of Personal Selling & Sales Management, 17 (4), 61-74.

Mahajan, Jayashree and Gilbert A. Churchill Jr. (1990), "Alternative Approaches for Investigating Contingency-Based Organizational Predictions in Personal Selling," International Journal of Research in Marketing, 7 (December), 149-69.

Marshall, Greg W., William C. Moncrief, and Felicia G. Lassk (1999), "The Current State of Sales Force Activities," Industrial Marketing Management, 28 (January), 87-98.

Mayr, Ernst (1969), Principles of Systematic Zoology. New York: McGraw-Hill.

McDonald, Malcolm, Tony Millman, and Beth Rogers (1997), "Key Account Management: Theory, Practice and Challenges," Journal of Marketing Management, 13 (November), 737-57.

McKelvey, Bill (1975), "Guidelines for the Empirical Classification of Organizations," Administrative Science Quarterly, 20 (December), 509-25.

Meyer, Alan D., Anne S. Tsui, and C.R. Hinings (1993), "Configurational Approaches to Organizational Analysis," Academy of Management Journal, 39 (6), 1175-95.

Miles, Raymond E. and Charles C. Snow (1978), Organizational Strategy, Structure and Process. New York: McGraw-Hill.

Milligan, Glenn W. (1996), "Clustering Validation: Results and Implications for Applied Analyses," in Clustering and Classification, P. Arabie, L.J. Hubert, and G. De Soete, eds. River Edge, NJ: World Scientific, 341-75.

--- and Martha C. Cooper (1985), "An Examination of Procedures for Determining the Number of Clusters in a Data Set," Psychometrika, 50 (2), 159-79.

--- and --- (1987), "Methodology Review: Clustering Methods," Applied Psychological Measurement, 11(4),329-54.

Millman, Tony F. (1996), "Global Key Account Management and Systems Selling," International Business Review, 5 (6), 631-45.

--- (1999), "From National Account Management to Global Account Management in Business-to-Business Markets," Thexis, 16 (4), 2-9.

---and Kevin Wilson(1989),"Processual Issues in Key Account Management: Underpinning the Customer-Facing Organization," Journal of Business & Industrial Marketing, 14 (4),328-37.

Moncrief, William C., III (1986), "Selling Activity and Sales Position Taxonomies for Industrial Sales forces," Journal of Marketing Research, 23 (August), 261-70.

Montgomery, David B. and George S. Yip (2000), "The Challenge of Global Customer Management," Marketing Management, 9 (4), 22-29.

Moon, Mark A. and Gary M. Armstrong (1994), "Selling Teams: A Conceptual Framework and Research Agenda," Journal of Personal Selling & Sales Management, 14 (1), 17-41.

--- and Susan F. Gupta (1997), "Examining the Formation of Selling Centers: A Conceptual Framework," Journal of Personal Selling & Sales Management, 17 (2), 31-41.

Moorman, Christine, Rohit Deshpande, and Gerald Zaltman (1993), "Factors Affecting Trust in Market Research Relationships," Journal of Marketing, 57 (January), 81-101.

Morgan, Robert M. and Shelby Hunt (1999), "Relationship-Based Competitive Advantage: The Role of Relationship Marketing in Marketing Strategy," Journal of Business Research, 46 (November), 281-90.

Murphy, K. and C. Davidshofer (1988), Psychological Testing: Principles and Applications. Englewood Cliffs, NJ: Prentice Hall.

Napolitano, Lisa (1997), "Customer-Supplier Partnering: A Strategy Whose Time Has Come," Journal of Personal Selling & Sales Management, 17 (4), 1-8.

Narus, James A. and James C. Anderson (1995), "Using Teams to Manage Collaborative Relationships in Business Markets," Journal of Business-to-Business Marketing, 2 (3), 17-46.

Nunnally, Jum C. (1967), Psychometric Theory, 1st ed. New York: McGraw-Hill.

--- (1978), Psychometric Theory, 2d ed. New York: McGraw-Hill.

Olson, Eric M., Orville C. Walker, and Robert W. Ruekert (1995), "Organizing for Effective New Product Development: The Moderating Role of Product Innovativeness," Journal of Marketing, 59 (January), 48-62.

Pardo, Catherine (1997), "Key Account Management in the Business to Business Field: The Key Account's Point of View," Journal of Personal Selling & Sales Management, 17 (4), 17-26.

--- (1999), "Key Account Management in the Business-to-Business Field: A French Overview," Journal of Business & Industrial Marketing, 14 (4), 276-90.

---, Robert Salle, and Robert Spencer (1995), "The Key Accountization of the Firm," Industrial Marketing Management, 22 (March), 123-34.

Pegram, Roger M. (1972), Selling and Servicing the National Account, Report No. 557. New York: The Conference Board.

Pelham, Alfred M. (1999), "Influence of Environment, Strategy, and Market Orientation on Performance in Small Manufacturing Firms," Journal of Business Research, 45 (1), 33-46.

Platzer, Linda Cardillo (1984), Managing National Accounts, Report No. 850. New York: The Conference Board.

Porter, Michael E. (1980), Competitive Strategy. New York: The Free Press.

Punj, Girish and David W. Stewart (1983), "Cluster Analysis in Marketing Research: Review and Suggestions for Application," Journal of Marketing Research, 20 (May), 134-48.

Rich, Philip (1992), "The Organizational Taxonomy: Definition and Design," Academy of Management Review, 17 (4), 758-81.

Rindfleisch, Aric (2000), "Organizational Trust and Interfirm Cooperation: An Examination of Horizontal Versus Vertical Alliances," Marketing Letters, 11 (1), 81-95.

Ruekert, Robert W., Orville C. Walker, and Kenneth J. Roering (1985), "The Organization of Marketing Activities: A Contingency Theory of Structure and Performance," Journal of Marketing, 49 (Winter), 13-25.

Sanchez, Julio C. (1993), "The Long and Thorny Way to an Organizational Taxonomy," Organization Studies, 14 (1), 73-92.

Sarle, Warren S. (1983), Cubic Clustering Criterion, SAS Technical Report A-108. Cary, NC: SAS Institute Publishing.

Sengupta, Sanjit, Robert E. Krapfel, and Michael A. Pusateri (1997a), "The Strategic Sales Force," Marketing Management, 6 (Summer), 29-34.

---, ---, and --- (1997b), "Switching Costs in Key Account Relationships," Journal of Personal Selling & Sales Management, 17 (4), 9-16.

Shapiro, Benson P. and Rowland T. Moriarty (1980), "National Account Management," Marketing Science Institute Working Paper No. 80-104. Cambridge, MA: Marketing Science Institute.

--- and --- (1982), "National Account Management: Emerging Insights," Marketing Science Institute Working Paper No. 82-100. Cambridge, MA: Marketing Science Institute.

--- and --- (1984a), "Organizing the National Account Force," Marketing Science Institute Working Paper No. 84-101. Cambridge, MA: Marketing Science Institute.

--- and --- (1984b), "Support Systems for National Account Management Programs," Marketing Science Institute Working Paper No. 84-102. Cambridge, MA: Marketing Science Institute.

--- and John Wyman (1981), "New Ways to Reach Your Customers," Harvard Business Review, 59 (July/August), 103-10.

Sharma, Arun (1997), "Who Prefers Key Account Management Programs? An Investigation of Business Buying Behavior and Buying Firm Characteristics," Journal of Personal Selling & Sales Management, 17 (4), 27-39.

Smith, J. Brock and Donald W. Barclay (1990), "Theoretical Perspectives on Selling Center Research," in Marketing Theory and Applications, David Lichtenthal et al., eds. Chicago: American Marketing Association, 5-11.

--- and --- (1993), "Team Selling Effectiveness: A Small Group Perspective," Journal of Business-to-Business Marketing, 1 (2), 3-32.

Sneath, Peter H. and Robert R. Sokal (1973), Numerical Taxonomy: The Principles and Practice of Numerical Classification. San Francisco: W.H. Freeman.

Spekman, Robert E. and Wesley J. Johnston (1986), "Relation-ship Management: Managing the Selling and the Buying Interface," Journal of Business Research, 14 (December), 519-31.

Steenkamp, Jan-Benedict E.M. and Hans Baumgartner (1998), "Assessing Measurement Invariance in Cross-National Consumer Research," Journal of Consumer Research, 25 (June), 78-90.

Stevenson, Thomas H. (1980), "Classifying a Customer as a National Account," Industrial Marketing Management, 9 (April), 133-36.

--- (1981), "Payoffs from National Account Management," Industrial Marketing Management, 10 (April), 119-24.

Stump, Rodney L. (1995), "Antecedents of Purchasing Concentration: A Transaction Cost Explanation," Journal of Business Research, 34 (October), 145-57.

Tice, Thomas E. (1997), "Managing Compensation Caps in Key Accounts," Journal of Personal Selling & Sales Management, 17 (4), 41-47.

Ward, J.H., Jr. (1963), "Hierarchical Grouping to Optimize an Objective Function," Journal of the American Statistical Association, 58, 236-44.

Weeks, William A. and Carl G. Stevens (1997), "National Account Management Sales Training and Directions for Improvement," Industrial Marketing Management, 26 (September), 423-31.

Weilbaker, Dan C. and William A. Weeks (1997), "The Evolution of National Account Management: A Literature Perspective," Journal of Personal Selling & Sales Management, 17 (4), 49-59.

Weitz, Barton A. and Kevin D. Bradford (1999), "Personal Selling and Sales Management: A Relationship Marketing Perspective," Journal of the Academy of Marketing Science, 27 (2), 241-54.

Workman, John P., Jr., Christian Homburg, and Kjell Gruner (1998), "Marketing Organization: An Integrative Framework of Dimensions and Determinants," Journal of Marketing, 62 (July), 21-41.

Wotruba, Thomas R. (1991), "The Evolution of Personal Selling," Journal of Personal Selling & Sales Management, 11 (3), 1-12.

--- and Stephen B. Castleberry (1993), "Job Analysis and Hiring Practices for National Account Marketing Positions," Journal of Personal Selling & Sales Management, 13 (3), 49-65.

Yip, George S. and Thammy L. Madsen (1996), "Global Account Management: The New Frontier in Relationship Marketing," International Marketing Review, 13 (3), 24-42.

Legend for chart:
A = Construct
B = Items
C = Composite Reliability/Coefficient Alpha


A  Activity intensity (reflective scale, scored on a seven-point
   scale with anchors 1 = "not more than for average accounts" and
   7 = "far more than for average accounts")
B  Compared to average accounts, to what extent do you do MORE in
   these areas for key accounts?
   • Product-related activities (e.g., product adaptation, new
   product development, technology exchange)
   • Service-related activities (e.g., training, advice,
   troubleshooting, guarantees)
   • Price-related activities (e.g., special pricing terms,
   corporatewide price terms, offering of financing solutions,
   revelation of own cost structure)
   • Distribution and logistics activities (e.g., logistics and
   production processes, quality programs, placement of own
   employees in account's facilities, taking over business processes
   from customer)
   • Information sharing (e.g., sharing of strategy and market
   research, joint production plans, adaptation of information
   systems, access to top management)
   • Promotion activities to final customers (e.g., joint
   advertising and promotion programs to help the account sell your
   products)
C  .75/.71

A  Activity proactiveness (formative scale, scored on a seven-point
   scale with anchors 1 = "not more than for average changes" and 7
   = "far more than for average changes")
B  Do the activities in these areas derive more from customer
   initiative or more from your own initiative? (Items equivalent to
   activity intensity.)
C

A  Top-management involvement (reflective scale, scored on a
   seven-point scale with anchors 1 = "strongly disagree" and 7 =
   "strongly agree")
B  Within our organization ...
   • even small matters related to key accounts have to be
   referred to someone higher up for a final decision.
   • very few decisions related to key accounts are made
   without the involvement of senior managers.
   • top management often deals with key account management.
C  .64/.62

A  Use of teams (reflective scale, scored on a seven-point scale
   with anchors 1 = "strongly disagree" and 7 = "strongly agree")
B  Within our organization ...
   • when there is a problem related to our key account
   relationships, a group is brought in to solve it.
   • key account-related decisions are made by teams.
   • we have teams that plan and coordinate activities for key
   accounts.
C  .85/.82

A  Selling center esprit de corps (adapted from Jaworski and Kohli
   1993; reflective scale, scored on a seven-point scale with
   anchors 1 = "strongly disagree" and 7 = "strongly agree")
B  People involved in the management of a key account ...
   • are genuinely concerned about the needs and problems of
   each other.
   • have a team spirit which pervades all ranks involved.
   • feel like they are part of a big family.
   • feel they are "in it together."
   • lack an "esprit de corps." (R)*
   • view themselves as independent individuals who have to
   tolerate others around them. (R)*
C  .92/.90

A  Access to marketing and sales resources (reflective scale, scored
   on a seven-point scale with anchors 1 = "very difficult" and 7 =
   "very easy")
B  How easy is it for the key account coordinator to obtain needed
   contributions for key accounts from these groups?
   • Field sales
   • Customer service
   • Product management
C  .75/.69

A  Access to nonmarketing and nonsales resources (reflective scale,
   scored on a seven-point scale with anchors 1 = "very difficult"
   and 7 = "very easy")
B  How easy is it for the key account coordinator to obtain needed
   contributions for key accounts from these groups?
   • Research and development
   • Manufacturing
   • Logistics
   • Finance/accounting
   • Information technology
   • General management
C  .85/.81

A  Approach formalization  (reflective scale, scored on a
   seven-point scale with anchors 1 = "strongly disagree" and 7 =
   "strongly agree")
B  Please indicate the extent to which you agree with the following
   statements:
   • We have established criteria for selecting key accounts.
   • Within our organization, formal internal communication
   channels are followed when working on key accounts.
   • To coordinate the parts of our organization working with
   key accounts, standard operating procedures have been
   established.
   • We have put a lot of thought into developing guidelines
   for working with our key accounts.
C  .87/.84

A  KAM effectiveness (reflective scale, scored on a seven-point
   scale with anchors 1 = "very poor," 4 = "about the same," and 7 =
   "excellent")
B  Compared to your average accounts, how does your organization
   perform with key accounts with respect to ...
   • achieving mutual trust?
   • achieving information sharing?
   • achieving a reputation of fairness?
   • achieving investments into the relationship?
   • maintaining long-term relationships?
   • reducing conflicts?
   • meeting sales targets and objectives?
   • making sales of those products with the highest margins?*
   • making sales from multiple product divisions?*
C  .88/.85

A  Performance in the market (reflective scale, scored on a
   seven-point scale with anchors 1 = "very poor," 4 = "about the
   same," and 7 = "excellent")
B  Relative to your competitors, how has your organization, over the
   last three years, performed with respect to ...
   • achieving customer satisfaction?
   • providing value for customers?
   • attaining desired growth?
   • securing desired market share?
   • successfully introducing new products?
   • keeping current customers?
   • attracting new customers?
C  .88/.85

A  Adaptiveness  (reflective scale, scored on a seven-point scale
   with anchors 1 = "not more than for average accounts" and 7 =
   "far more than for average accounts")
B  Relative to your competitors, how has your organization, over the
   last three years, performed with respect to ...
   • adapting to changes in the business environment of your
   company?
   • adapting to changes in competitors' marketing strategies?
   • adapting your products quickly to the changing needs of
   customers?
   • reacting quickly to new market threats?
   • exploiting quickly new market opportunities?
C  .86/.84

A  Profitability (interval item with ten levels of variable provided)
B  What was your company's average pre-tax profit margin over the
   last three years? 1 = negative; 2 = 0%-2%, 3 = 2%-4%, 4 = 4%-6%,
   5 = 6%-8%, 6 = 8%-10%, 7 = 10%-12%, 8 = 12%-16%, 9 = 16%-20%,
   10 = more than 20%
C

A  Competitive intensity (adapted from Jaworski and Kohli 1993;
   reflective scale, scored on a seven-point scale with anchors 1 =
   "strongly disagree" and 7 = "strongly agree")
B  Please indicate the extent to which you agree with the following
   statements:
   • Competition in our industry is cutthroat.
   • There are many "promotion wars" in our industry.
   • Anything that one competitor can offer, others can match
   readily.
   • Price competition is a hallmark of our industry.
   • One hears of a new competitive move almost every day.
   • Our competitors are relatively weak. (R)*
C  .82/.81

A  Market dynamism (adapted from Jaworski and Kohli 1993; reflective
   scale, scored on a seven-point scale with anchors 1 = "strongly
   disagree" and 7 = "strongly agree")
B  Please indicate the extent to which you agree with the following
   statements:
   • In our kind of business, customers' product preferences
   change quite a bit over time.
   • Our customers tend to look for new products all the time.
   • We are witnessing demand for our products and services
   from customers who never bought them before.
   • New customers tend to have product-related needs that are
   different from those of our existing customers.
   • We cater to many of the same customers that we used to in
   the past. (R)*
C  .65/.61

*Items not kept after the item purification process.

~~~~~~~~

By Christian Homburg; John P. Workman Jr. and Ove Jensen

Christian Homburg is Professor of Marketing and Director of the Institute for Market-Oriented Management, University of Mannheim. John P. Workman Jr. is Associate Professor of Marketing, College of Business Administration, Creighton University. Ove Jensen is a partner and managing director, Prof. Homburg & Partners. The research reported in this article was supported by funding from the Marketing Science Institute. The authors acknowledge the research assistance provided by Christian Johann, Jan Loewner, Andrea Model, Christine Prauschke, and Michaela Vogel in Germany and Brenda Gerhardt and Anurag Aerora in the United States.

A Customer Lifetime Value Framework for Customer Selection and Resource Allocation Strategy 

The authors evaluate the usefulness of customer lifetime value (CLV) as a metric for customer selection and marketing resource allocation by developing a dynamic framework that enables managers to maintain or improve customer relationships proactively through marketing contacts across various channels and to maximize CLV simultaneously. The authors show that marketing contacts across various channels influence CLV nonlinearly. Customers who are selected on the basis of their lifetime value provide higher profits in future periods than do customers selected on the basis of several other customer-based metrics. The analyses suggest that there is potential for improved profits when managers design resource allocation rules that maximize CLV. Managers can use the authors' framework to allocate marketing resources efficiently across customers and channels of communication.

Customer lifetime value (CLV) is rapidly gaining acceptance as a metric to acquire, grow, and retain the "right" customers in customer relationship management (CRM). However, many companies do not use CLV measurements judiciously. Either they work with undesirable customers to begin with, or they do not know how to customize the customer's experience to create the highest value (Thompson 2001). The challenge that most marketing managers currently face is to achieve convergence between marketing actions (e.g., contacts across various channels) and CRM. Specifically, they need to take all the data they have collected about customers and integrate them with how the firm interacts with its customers. In the academic literature, Berger and colleagues (2002) support the allocation of resources to maximize the value of the customer base, and they strongly argue that such resource allocation models are needed.

Some researchers have recommended CLV as a metric for selecting customers and designing marketing programs (Reinartz and Kumar 2003; Rust, Zeithaml, and Lemon 2004). However, there is no empirical evidence as to the usefulness of CLV compared with that of other customer-based metrics. Table 1 compares the CLV framework proposed in this study with the existing literature on CLV and database marketing. A comparison of the studies listed in Table 1 shows that most of the previous studies provide guidelines for calculating CLV and return on investment at the aggregate level. Some recent studies (Reinartz and Kumar 2003; Rust, Zeithaml, and Lemon 2004) provide empirical evidence for the existence of a relationship between marketing actions and CLV at the aggregate level. However, as Berger and colleagues (2002) state, none of the studies has proposed and tested a framework that provides rules for resource allocation across various channels of communication for each individual customer and across customers.

On the basis of the comparisons in Table 1, we summarize the significant contributions of our study as follows: We provide a framework for measuring CLV that links the influence of communications across various channels on CLV. We also evaluate the usefulness of CLV as a metric for customer selection and develop a framework for marketing resource allocation that maximizes CLV. Given the assumed link between CLV and firm profitability (Hogan et al. 2002), these are important issues.

In this study, we use customer data from a large business-to-business (B2B) manufacturer to illustrate the proposed framework empirically. The customer database of the organization focuses on B2B customers. Our analyses show that marketing communications across various channels influence CLV nonlinearly. The results from our analyses suggest that customers selected on the basis of CLV provide higher profits than do customers selected on the basis of other widely used CRM metrics. In addition, there is the potential for substantial improvement in profits when managers design resource allocation rules that maximize CLV.

In the next section, we develop the framework for the measurement and maximization of CLV. We then propose hypotheses about the influence of supplier-specific factors and customer characteristics on the various CLV components. In the subsequent section, we explain the models and data we used to estimate CLV. We then discuss the results from our analyses and explain the comparison of CLV with other metrics for customer selection. Specifically, we compare the aggregate profits provided by high-CLV customers with those of customers who score high on several other customer-based metrics. In the section "Resource Allocation Strategy," we provide details on allocating resources that maximize CLV. Our objective there is to evaluate the extent to which CLV, and thus profits, can be increased by allocating marketing resources across channels of contact for each customer so as to maximize his or her respective CLVs. Finally, we derive implications based on the results, discuss the limitations of our study, and identify areas for further research.

The various components of CLV include purchase frequency, contribution margin, and marketing costs (however, the various CLV components can vary depending on the industry). Some of the antecedents of purchase frequency and contribution margin (e.g., marketing communications) are under management's control and affect the variable costs of managing customers. We use these antecedents to maximize CLV.

Objective Function: CLV

Typically, CLV is a function of the predicted contribution margin, the propensity for a customer to continue in a relationship (customer retention), and the marketing resources allocated to the customer. In general, CLV can be calculated as follows:

[Multiple line equation(s) cannot be represented in ASCII text]

where

i = customer index, t = time index, n = forecast horizon, and r = discount rate.

In contractual settings, managers are interested in predicting customer retention, or the likelihood of a customer staying in or terminating a relationship. However, in noncontractual settings, the focus is more on predicting future customer activity because there is always a chance that the customer will purchase in the future. Therefore, managers who calculate CLV in noncontractual settings are interested in predicting future customer activity and the predicted contribution margin from each customer. Previous researchers have used the variable P(Alive), which represents the probability that a customer is alive (and thus exhibits purchase activity) given his or her previous purchase behavior (Reinartz and Kumar 2000), to predict future customer activity in noncontractual settings. However, the measure assumes that when a customer terminates a relationship, he or she does not return to the supplier. This is also called the "lost-for-good" scenario ((Rust, Zeithaml, and Lemon 2004). If a customer is won back after termination, the company treats the customer as a new customer and ignores its history with the customer.

Another method for predicting future customer activity is to predict the frequency of a customer's purchases given his or her previous purchases. The assumption underlying this framework is that customers are most likely to reduce their frequency of purchase before terminating a relationship. This assumption is in accordance with theories about the different phases in a relationship and relationship life cycles (Dwyer, Schurr, and Oh 1987; Jap 2001). In addition, such a methodology enables a customer to return to the supplier after a temporary dormancy in a relationship. Thus, in this framework, we measure CLV by predicting the purchase pattern (purchase frequency or interpurchase times) over a reasonable period. This is also called the "always-a-share" scenario.. The lost-for-good approach is questionable because it systematically understates CLV (Rust, Zeithaml, and Lemon 2004). Thus, we use the always-a-share approach in this study. Given predictions of contribution margin, purchase frequency, and variable costs, the CLV function we use can be represented as follows:

( 2) [Multiple line equation(s) cannot be represented in ASCII text]

where

CLV[subi] = lifetime value of customer i;

CM[subi,y] = predicted contribution margin from customer i (computed from a contribution margin model) in purchase occasion y, measured in dollars;

r = discount rate for money (set at 15% annual rate in our study);

c[subi,m,l] = unit marketing cost for customer i in channel m in year l (the formulation of CLV does not change if l is used to represent periods other than one year);

x[subi,m,l] = number of contacts to customer i in channel m in year l;

frequency[subi] = predicted purchase frequency for customer i;

n = number of years to forecast; and

T[subi] = predicted number of purchases made by customer i until the end of the planning period.

In addition to accurate measurement of CLV for each customer, our objective is to allocate resources so as to maximize CLV. Thus, we model the purchase frequency and contribution margin of customers as a function of marketing resource variables such as channel contact. We then use the customer responsiveness to marketing actions, obtained from the purchase frequency and contribution margin models, to develop resource allocation strategies that maximize CLV. In summary, the objective is to identify the resource allocation rules across various channels of communication for each individual customer such that the respective CLVs (as provided in Equation 2) are maximized.( n1) Our objective function is subject to the following constraints: frequency > 0 ∀ i, t, and x[subi,m,l] ≥ 0 ∀ i, m, l.

Discounting contribution margin. We first focus on the discounting of contribution margin over a period of time. Assume that it is currently year l = 1 and that we need to forecast the contribution margin from each customer for the next n years (i.e., until l + n). It is possible that a customer makes several purchases in a given year. Berger and Nasr (1998, Equation 2) and Rust, Zeithaml, and Lemon (2004) provide guidelines for discounting contribution margin from customers when there is more than one purchase occasion (y) per year. In this approach, the discount rate from a customer is scaled according to his or her frequency of purchase (as is shown in Equation 2). For example, consider when the planning horizon is one year and the frequency of purchases is two times (frequency = 2). The first purchase occasion (y = 1) occurs after 6 months; therefore, y/frequency = .5 (in other words, we use the square root of the discount rate). The second purchase occasion (y = 2) occurs after 12 months; therefore, y/frequency = 1.

Discounting cost allocations. The discounting of cost allocations is straightforward if we assume that there is a yearly allocation of resources (as is the case in most organizations) and that the cost allocation occurs at the beginning of the year (the present period). Thus, the cost allocation in the first year need not be discounted, the cost allocation in the second year needs to be discounted for one year, and so on. Thus, we raise the denominator in the cost function calculation to current year - 1 ((i.e., l - 1)).

Discussion of model constraints. The constraints ensure the nonnegativity of the predicted purchase frequency and communication levels for each customer i during period l.

Purchase Frequency

An objective of relationship marketing is to ensure future purchase activity. Purchase frequency is also a component of our CLV calculation. Therefore, as the basis for selecting antecedents to predict purchase frequency, we use the commitment trust theory of relationship marketing (Morgan and Hunt 1994) as well as previous research in customer equity and CLV (Bolton, Lemon, and Verhoef 2004; Bowman and Narayandas 2001; Reinartz and Kumar 2003; Rust, Zeithaml, and Lemon 2004) and channel communications (Grewal, Corner, and Mehta 2001; Mohr and Nevin 1990; Morgan and Hunt 1994; Rindfleisch and Heide 1997).

The overall theoretical framework that we used is provided in Figure 1. We summarize the antecedents of purchase frequency, their operationalization, expected effects, and the rationale for our expectations in Table 2. Next, we provide a detailed discussion for a few of the hypotheses that are unique to our study.

Supplier-Specific Factors: Channel Communications

In this study, we classify channels of communication into the following contact modes: rich (e.g., face-to-face, trading event meetings), standardized (e.g., direct mail, telephone), and Web based (Mohr and Nevin 1990). Although we expect the relationships between different channels of communication and predicted customer activity to be similar, we need to analyze customer responses separately across different channels because the costs of serving customers across different channels are different, and customers might exhibit different responsiveness across the various channels. The costs of communication in each channel can influence managers frequency of communication in each channel.

Frequency of rich and standardized modes. Face-to-face communications and trading event meetings are the richest and most direct mode of communication possible among channel members (Mohr and Nevin 1990). Relational customers tend to have high commitment and trust with their suppliers, which results in less uncertainty, more cooperation, and less complexity in their relationships than in those of transactional customers (Morgan and Hunt 1994). Rich modes of communication are preferred to standardized modes when issues in the channel structure are complex and when there is a high degree of uncertainty in the relationship. Rich modes of communication are also effective in converting transactional customers to relational ones (Ganesan 1994).

Direct mail and telephone communication are the most standardized and cost-effective modes of individual-level communication available to an organization. Standardized modes are also the most cost-effective method for identifying customers who are interested in an organization's current promotion (Shepard 2001). For transactional customers, direct mail can be used in combination with telephone sales to generate interest in products while simultaneously improving the return on investment (Nash 1993). For relational customers, direct mail serves to maintain commitment and trust by communicating relationship benefits (Morgan and Hunt 1994) and to inform the best customers about new product offerings. Therefore, although the purpose of standardized communication may be different for transactional customers than for relational ones, we expect that the marginal response for increased frequency is the same across segments.

However, it has been proposed that too much communication causes a relationship to be dysfunctional (Fournier, Dobscha, and Mick 1997). In addition, the marginal response to a higher level of rich modes of communication need not always be higher; sometimes, it even can be negative. Although the utility of marketing contacts is not questioned, too much contact can overload buyers and have dysfunctional consequences (e.g., ubiquitous junk mail). Thus:

H[sub1]: An inverted U-shaped relationship exists between the frequency of rich and standardized modes of communication and a customer's predicted purchase frequency.

Intercontact time. Following the theory that leads to H1, we expect that there exists an optimal level of intercontact time between suppliers and buyers. Higher levels of previous communications lead to trust with the supplier and act as glue that holds together a communication channel (Morgan and Hunt 1994). However, too much communication may be dysfunctional. In addition, the marginal utility of an additional piece of information from a supplier firm in a short period is low. To maximize the effect of each contact, supplier firms need to pace their communication schedule to suit customer needs. Thus:

H[sub2]: An inverted U-shaped relationship exists between intercontact time and a customer's predicted purchase frequency.

Customer Characteristics: Customer Involvement

Bidirectional communication. Research on channel communications shows that highly relational channel structures are associated with large bidirectional communications among channel members (Mohr and Nevin 1990). Although customer-initiated contacts are associated primarily with complaints in business-to-consumer settings, the same is not necessarily true for B2B settings. In a B2B setting, customers can initiate contacts with suppliers for several reasons, such as if they have new needs that the supplier may be able to fulfill, if they want the supplier to conduct training programs at the customer's site (Cannon and Homburg 2001), or if the supplier invites the customer to participate in new product development sessions. On most occasions, bidirectional communication in channels strengthens a relationship, indicates customer involvement, and increases interdependence among channel members (Ganesan 1994; Mohr and Nevin 1990). Thus:

H[sub3]: The higher the level of bidirectional communication, the higher is a customer's predicted purchase frequency.

Frequency of Web-based contacts. In this study, we analyze Web-based contacts separately from other channels of communication because Web-based communication is customer initiated (i.e., a passive mode of operation for the supplier). However, there are several advantages to tracking Web-based contacts in a B2B setting. First, Web-based communication between buyers and suppliers is the most cost-effective method of communication. Second, Web-based contacts from the buyers provide some important signals to the supplier about the buyer's relationship orientation. Grewal, Corner, and Mehta (2001) find that organizations enter and actively participate in electronic markets if their motivation is to improve efficiency in transactions. In addition, participation in electronic markets (or use of Web-based initiatives) improves transaction effectiveness and efficiency (Rindfleisch and Heide 1997). Efficiency of communication and transactions among channel members is associated with a relational structure and higher customer involvement (Mohr and Nevin 1990; Sheth and Parvatiyar 1995). Thus:

H[sub4]: The higher the number of Web-based contacts from a customer, the higher is the customer's predicted frequency of purchase.

Contribution Margin

The antecedents that we adopt to predict contribution margin are based on findings from previous research on antecedents of customer revenue (Niraj, Gupta, and Narasimhan 2001) and purchase quantity (Gupta 1988; Tellis and Zufryden 1995). As with purchase frequency, we classify the antecedents of contribution margin as supplier-specific factors (total marketing efforts) and customer characteristics (lagged contribution margin and purchase quantity). We use size of an establishment and industry category as covariates in our model. In Table 2, we provide a description of the antecedents that we propose influence customer purchase frequency and contribution margin. In the database, we also provide the operationalization of the antecedents, their expected effects, and the rationale for our expectations based on previous research. Because all the antecedents we use in our contribution margin model are based on previous research and findings, we do not discuss the hypotheses in detail. In Table 2, we provide a description of the covariates we use as control variables in the purchase frequency and contribution margin models.

Model Development

To predict CLV, we need a stochastic model to predict each customer's purchase frequency and a panel-data model that predicts contribution margin. In this study, we assume that the amount a customer spends is independent of purchase timing. This is a rather restrictive assumption for frequently bought consumer goods (Tellis and Zufryden 1995). However, in our product category, we find that the correlation between purchase frequency and contribution margin is not significant.

Purchase Frequency

We model a customer's purchase frequency using the generalized gamma model of interpurchase timing that Allenby, Leone, and Jen (1999) developed. The generalized gamma model also accommodates the commonly used exponential distribution for interpurchase times (Reinartz and Kumar 2003). The likelihood function for the purchase frequency model is given as follows:

( 3) [Multiple line equation(s) cannot be represented in ASCII text]

where

f(t[subij]|α, λ[subi], γ) = the density function for the generalized gamma distribution (i.e., the probability of the jth purchase for customer i occurring at period t, given α, λ[subi]; γ);

S(t[subij]|α, λ[subi], γ) = the survival function for the generalized gamma distribution (i.e., the probability of the jth purchase for customer i occurring at a given period is greater than t, given that the jth purchase has not occurred until time t, given α λ[subi], γ);

c[subij] = the censoring indicator, where c[subij] = 1 if the jth interpurchase time for the ith customer is not right-censored, and c[subij] = 0 if the jth interpurchase time for the ith customer is right-censored;

φ[subijk] = the probability of observation j for the ith customer belonging to subgroup k; and

α λ[subi], γ = the parameters of the generalized gamma distribution.

Because we use a generalized gamma distribution to model interpurchase time and the likelihood function in Equation 3, the expected time until next purchase is given as follows:

( 4) [Multiple line equation(s) cannot be represented in ASCII text]

The ratio of 12 (because we use months as the unit of analysis) to the expected time until next purchase (which we obtain by modeling a generalized gamma distribution on the interpurchase times, as is shown in the work of Allenby, Leone, and Jen [1999]) gives the predicted purchase frequency. The parameters α and γ establish the shape of the interpurchase time distribution, and λ[sub1] is the individual-specific purchase rate parameter. We assume that the population consists of k subgroups, and φ[subik] provides the mass point (i.e., weight) for each subgroup. We model the probability of a customer belonging to each subgroup φ[subik] as a probit function of the antecedents and covariates of purchase frequency. Specifically, we represent the link function as φ[subik] = f(x[subij]β[subi]), where x[subij] represents the antecedents and covariates of purchase frequency for customer i in purchase occasion j, and β[subi] represents the customer-specific response coefficients.

Our model framework, as presented in Equation 3, resembles a hierarchical Bayes formulation of the concomitant continuous mixture model. To address the issue of endogeneity, we use the one-period lagged value for all the antecedents and covariates in our analysis (Villas-Boas and Winer 1999). However, to account for any extraneous factors, we also use the log of the lagged interpurchase.( n2) The specification of the model enables us to estimate individual customer-level coefficients for the influence of the covariates on the probability of a customer belonging to a particular subgroup and thus interpurchase times.

Contribution Margin

We model the contribution margin from a customer using panel-data regression methodologies. We needed to address endogeneity issues while using lagged contribution margin as an independent variable in our model. In panel-data models with lagged dependent variables, the endogeneity in formulation can be alleviated with a one-period difference in the dependent variable and a two-period lagged dependent variable as an independent variable (Baltagi 1998). We use the growth in contribution margin from period t - 1 to t as the dependent variable and the contribution margin in period t - 2 as an independent variable. The other independent variables we used are specific to period t - 1 ((this also accommodates the issue of endogeneity). The independent variables in the contribution margin model are thus lagged contribution margin, lagged total quantity purchased, lagged firm size, industry category, and lagged total marketing efforts. Thus, the contribution margin model is

( 5) ΔCM[subi,t] = β[sub0] + β[sub1]CM[subi,t - 2] + β[sub2]Quantity[subi,t - 1] + β[sub3]Size[subi,t - 1] + Σ[subj]β[subj]Industry[subj] + β[sub4]Totmark[subi, t - 1] + e[subi,t],

where

ΔCM[subi,t] = difference in contribution margin from period t 1 to period t for customer i, measured in dollars;

Size[subi, t = 1 = firm size for customer i in period t - 1,, measured as number of employees;

Industry = indicator variable for industry category of the customer firm;

Totmark[subi, t = 1 = total number of contacts made to customer i in period t - 1;

Quantity[subi, t = 1] = total quantity of products bought by customer i in period t - 1;

e[subi,t] = error term;

i = index for the customer; and

t = index for time.

Data

We used data from a large multinational computer hardware (servers, workstations, and personal computers) and software (integration and application) manufacturer for the empirical application of our framework. The company's database focuses on business customers. The product categories in the database represent different areas of high-technology products. In addition, for these product categories, it is the choice of the buyer and seller to develop their relationships, and there are significant benefits for both parties to maintain a long-standing relationship. The choice of vendors for the products is normally made after much deliberation by the buyer firm. Even though the firm's products are durable goods, they require constant maintenance and frequent upgrades, which provides the variance required in modeling customer response. For our analyses, we used two cohorts of customers: Cohorts 1 and 2. We assigned customers to Cohort 1 (Cohort 2) if their first purchase with the manufacturer occurred in the first quarter of 1997 (first quarter of 1998). In our samples, we removed customers who had missing values for either rich or standardized modes of communication. We also restricted our sample to customers who had made at least five purchases. Overall, we removed 20% of the original cohort of customers for our analyses, which resulted in an effective sample size of 1316 and 873 observations for customers in Cohorts 1 and 2, respectively. The interpurchase time for customers in Cohort 1 ranges from 1.5 to 23 months; for Cohort 2, it ranges from 1 month to 20 months.

Purchase frequency model. We used each observed purchase for a customer as an observation in the purchase frequency model. For Cohort 1, we selected customers who made their first purchase in the first quarter of 1997. For each customer, we omitted the first observed purchase in our analysis sample because the first observed purchase is restricted to be within three months for all customers in the cohort, and theoretically the customer retention phase begins after the first purchase. The antecedents and covariates we used can be classified as cumulative and current effects. The cumulative effects antecedents include cross-buying and upgrading, and their values represent the total number of different products (for cross-buying) or upgrades that the customer has purchased since the first purchase until the current observed purchase.

The current-effects antecedents include bidirectional communication; returns; relationship benefits; frequency of rich, standardized, and Web-based contacts; and intercontact time. The covariates in the purchase frequency model (type of product purchased) can be classified as current effects. We calculated the current-effects antecedents and covariates on the basis of the activities of the customer or the supplier (in the case of channel communications) between the previous observed purchase (j - 1)) and the current observed purchase (j). To assess the inverted U-shaped relationships, we used a quadratic conversion (including the square covariate term in Equation 3) of the respective antecedent. For all customers, we used the interpurchase times until the end of 2000 as our calibration sample. We used the 2001 data as a holdout sample and to compare strategies. All the antecedents and covariates we used in our analyses are lagged variables. Specifically, for observed purchase j, the cumulative-effects antecedents represent the customer's activity since relationship initiation until observed purchase j - 1. Similarly, for observed purchase j, the current effects antecedents and covariates represent the customer's (or supplier s) activity between observed purchases j - 2 and j - 1.

Contribution margin model. To model contribution margin from a customer, we used the annual sales from various purchases of each customer. For customers in Cohort 1, we used the annual sales from each customer from 1997 to 2000. Given our model structure in Equation 5, there are two observations per customer in our analysis sample. Specifically, for each customer, for Observation 1 the dependent variable is the difference in contribution margin between 2000 and 1999, and the independent variables include the contribution in 1998, the firm size in 1999, the industry category of the customer, the total number of contacts made to the customer in 1999, and the total quantity of products purchased in 1999. Similarly, for Observation 2, the dependent variable is the difference in contribution margin between 1999 and 1998, and the independent variables include the contribution margin in 1997. As we stated previously, we used the 2001 data as a holdout sample and to compare strategies. The descriptive statistics for the data and the correlation matrix of the antecedents are provided in Table 3.

Purchase Frequency Model

As we discussed previously, we used an effective sample size of 1316 and 873 observations that belong to Cohort 1 (first purchase in 1997) and Cohort 2 (first purchase in 1998), respectively, for our analyses. We discuss the results from Cohort 1 in detail in the text. The results from Cohort 2 are quite similar to those of Cohort 1 and are provided along with the results for Cohort 1 in the corresponding tables. We censored our data set in 2000 and used the 2001 data as our validation (or holdout) sample in both cohorts. We estimated the purchase frequency model in Equation 3 using Markov chain Monte Carlo (MCMC) methods. The results from the purchase frequency model for Cohorts 1 and 2 are provided in Table 4 (for details on the model estimation, model selection, model performance, and support for hypotheses, see Appendix A). The coefficients of the antecedents reported in Table 4 are the means from the posterior samples of β[subi], and the signs for the coefficients represent their influence on a customer's purchase frequency. There are several insights that we derive from Table 4, which we discuss subsequently.

Model fit. The results show that the generalized gamma model with two subgroups provides a good fit to the data and is better than other models for modeling purchase frequency (log-likelihood for Model 3 = -2237.82, Akaike information criterion [AIC] = -4484, and Bayesian information criterion [BIC] = -4683). We also used a hazard model with the finite mixture framework (Kamakura and Russell 1989) to model purchase frequency, and we found that our proposed model fits the data better and has better predictive capabilities.

Distribution parameters.( n3) The mean expected purchase frequency for Subgroup 1 is 4.2 purchases in a year, and the mean expected purchase frequency for Subgroup 2 is 1.01 purchases in a year. Given the variation in expected purchase frequencies in each subgroup, we term Subgroup 1 the active state and Subgroup 2 the inactive state. The component masses for Subgroup 1 (φ[sub1]) and Subgroup 2 (φ[sub2]) are .54 and .46, respectively. This implies that we expect 54% of the customers to be active in the prediction window and 46% of the customers to be inactive in the prediction window.

Supplier-specific factors. Our analyses indicate that a supplier's contact strategy and provision of relationship benefits significantly affect a customer's predicted purchase frequency. We find that the frequency of contacts affects purchase frequency nonlinearly. Specifically, we find an inverted U-shaped relationship. This leads us to believe that there is an optimal level of marketing communication for each customer. A firm's increasing communication beyond a certain threshold may result in diminishing returns in terms of customer purchase frequency. This finding also provides the reasoning to determine the optimal level of resources that needs to be allocated across channels to maximize CLV in Phase 2.

The coefficients of the marketing contacts reveal a difference in the influence of various channels on customer purchase frequency. The coefficient of the quadratic term for rich modes of communication (-1.30 for Cohort 1) is higher than the coefficient of the quadratic term for standardized modes of communication (-.28 for Cohort 1). Thus, we can infer that the rate of diminishing returns (after exceeding the threshold) is much higher for the rich mode of communications than for standardized modes. Therefore, although the rich mode of communication is interactive and effective, firms should use it with great caution.

Customer characteristics. The results indicate that upgrading and cross-buying positively influence a customer's purchase frequency. This is in line with the findings of Reinartz and Kumar (2003), who also find that breadth of purchase positively affects a customer's duration in a profitable relationship. We also find that the higher the bidirectional communication between the customer and supplier, the higher is the customer's purchase frequency. Thus, in addition to timely communication from the supplier to the customer (Morgan and Hunt 1994), communication from a customer can be a good indicator of a customer's activity.

With respect to returns from a customer, our analysis suggests that managers need to exercise caution. We find support for an inverted U-shaped relationship between purchase frequency and returns. This indicates that customers who return products within a threshold are a good asset to the firm. The results highlight the importance of firms' recognizing the customers who establish contact through the online channel in their CRM strategies.

Contribution Margin Model

We estimated the contribution margin model with annual data from 1997 (t - 4)) to 2000 (t) for Cohort 1 and from 1998 to 2000 for Cohort 2.( n4) The revenue in 2001 (t + 1) acts as a holdout sample for Cohorts 1 and 2. The coefficients of the contribution margin model are provided in Table 5. The main insight from Table 5 is that the contribution margin model provides an adjusted R2 of .68 and thus can explain significant variation in contribution margin from customers. We derive several other insights from Table 5, which we discuss subsequently.

Supplier-specific factors. Total lagged marketing efforts contribute significantly to an explanation of variation in current contribution margin, which implies that a supplier's contact strategy affects both purchase frequency and contribution margin.

Customer characteristics. The two-period lagged contribution margin provides the highest contribution to an explanation of current-period contribution margin. In addition, lagged quantity of goods is significant in explaining variation in contribution margin. Firm size and industry category explain variation in current-period contribution margin. Among the various industry categories, firms in the financial services, technology, consumer packaged goods, and government industry categories provide, on average, a higher contribution margin than do firms in other industry categories. However, firms in the education industry provide a lower contribution margin than do firms in other industries.

In this section, we compare the customer selection capabilities of the following: CLV, our proposed metric; previous-period customer revenue (PCR), a simple metric; past customer value (PCV), which is widely considered a good predictor of future customer value; and customer lifetime duration (CLD), a forward-looking metric that is also used as a proxy for loyalty. (We also compared the customer selection capabilities of CLV with other customer-based metrics, such as share of wallet and recency, frequency, and monetary value. The results were similar to the comparison with PCR, PCV, and CLD.) In general, organizations in direct marketing situations rank-order their customers on the basis of a particular metric and prioritize their resources from best customers to worst customers on the basis of the rank order (Roberts and Berger 1999). Descriptions of the various customer-based selection metrics used in our analyses are provided in Appendix B.

Performance of the Customer Selection Metrics

To compare the performance of the four metrics, we rank-ordered customers from best to worst according to each metric and then compared the sales, costs, and profits from the top 5%, 10%, and 15% of customers. We used the data from the first 30 months to score and sort the customers on each metric, as do Reinartz and Kumar (2003). We then compared the actual sales, variable costs of communication, and profits for the top 5%, 10%, and 15% of customers from the censoring period (30 months) until the end of the observation window (48 months).( n5) To select customers for contact, in general organizations choose the top 5% to 15% of their customers, rank-ordered on the basis of a scoring metric. Selection of more customers to contact may not be feasible because of limited time and resources. Thus, to reflect industry practice, we compared the performance of our metric among the top 5%, 10%, and 15% of customers. The results are provided in Table 6, and the reported values are cell medians. We subsequently summarize the results from our comparison.

Overall, Table 6 shows that the proposed metric better identifies profitable customers than do other metrics we compared in the study, such as PCR, PCV, and CLD. On the basis of the 18-month prediction window, we expect the average net profits of customers selected from the top 5% using the proposed CLV metric to be $143,295 (after accounting for cost of goods sold [70%] and variable costs of communication), whereas the average net profits are $70,929, $130,785, and $106,389 for the top 5% of the customers selected on the basis of PCR, PCV, and CLD, respectively. These findings hold across all the percentage subgroups. The results provide substantial support for incorporating the responsiveness of each individual customer across various communication channels and for the usefulness of CLV as a metric for customer scoring and customer selection. Although the difference in profits from use of PCR, PCV, CLD, and CLV is, on average, approximately $40,000 for a customer in the top 5% sample, the difference in total profit across the top 5% to 15% of the entire customer base can easily yield more than $1 million.

Having evaluated the usefulness of using CLV for customer selection, we now describe our methodology for designing resource allocation strategies that maximize CLV. The framework also provides managers a tool for assessing return on marketing investments by identifying avenues for optimal resource allocation across channels of communication for each individual customer (and possibly across customers), so as to maximize CLV. The marketing literature has provided guidelines for optimal resource allocation in acquisition and retention decisions (Blattberg and Deighton 1996; Blattberg, Getz, and Thomas 2001), promotion expenditures (Berger and Bechwati 2001; Berger and Nasr 1998), and marketing actions when future brand switching is considered (Rust, Zeithaml, and Lemon 2004). These guidelines represent a significant step toward incorporation of long-term customer profitability effects into firm-level managerial decision making. However, the models provide less insight into decisions about how to manage individual customers in a way that accounts for the heterogeneity, and they do not provide a mechanism for dynamic updating of profitability assessment (Libai, Narayandas, and Humby 2002).

Our resource allocation algorithm uses Equation 2 as the objective function, and the purpose of the optimization is to find the level of contacts across various channels with each individual customer that would maximize CLV. Equation 2 is a function of predicted purchase frequency (based on Equation 4), predicted contribution margin (based on Equation 5), and marketing costs. We first estimated the responsiveness of customers (coefficients, β[subs] to marketing contacts from the purchase frequency model and the contribution margin model. To design a CLV-based resource allocation strategy, we held the coefficients constant and identified the level of covariates (i.e., level of channel contacts) for each customer that would maximize CLV. In summary, in Equation 2, the contacts made to a customer across various channels are under the supplier's control and thus can be used to maximize CLV, depending on the cost of each mode of communication and the responsiveness of the customer (in terms of both purchase frequency and contribution margin) to each channel of communication.( n6) In other words, with respect to marketing resources for a firm, the customer contact levels across different channels appear in the revenue and cost sides of Equation 2 and thus avoid the scope of corner solutions.

We used a genetic algorithm to derive the levels of contact desired for each individual customer that maximize CLV. Genetic algorithms (Goldberg 1989) are simulation-based, parallel-search algorithms that have been used in econometrics (Dorsey and Mayer 1995; Liang and Wong 2001) to obtain optimal solutions when the complexity of the optimization function tends to be intractable and multidimensional. In our study, support for a purchase frequency distribution with two subgroups led us to believe that the optimization surface is multimodal. In addition, we intended to allocate resources for each customer on the basis of individual responsiveness. These issues made our optimization problem extremely complex and intractable with traditional analytical methods. Thus, we resorted to a search algorithm to find the optimal resource allocation levels. In addition, the multimodal nature of the optimization surface (given the support for a mixture distribution for purchase frequency) motivated us to use a parallel-search technique, which is not susceptible to local minima (common in multimodal distributions) (Venkatesan, Krishnan, and Kumar 2004). Appendix C explains how we used a parallel-search technique for resource allocation purposes.

Aggregate Results

The total net present value of future profits from a resource allocation strategy that maximizes CLV (with the predicted contribution margin) is approximately $44 million. We also computed the total net present value of future profits when the organization uses its current resource allocation strategy. Specifically, for each customer, we maintained the resource allocation levels for the most recent year and calculated CLV over a three-year period. We find that, based on this status quo resource allocation strategy, the total net present value of future profits is approximately $24 million. Therefore, we find that a resource allocation strategy that maximizes CLV results in an increase in profits by approximately 83%. The total cost of communication (over three years), based on the resource allocation strategy that maximizes CLV, is approximately $1 million. The total cost of communication in the organization's current strategy is approximately $716,188. We find that the organization improves profits by increasing costs of serving customers (rather than cost of communication in the previous year) by 48%. The return on marketing communication to the organization, based on its current strategy, is approximately $34 million ($24 million/$716,188). With a communication strategy that maximizes CLV, the return on marketing communication to the organization is approximately $44 million ($44 million/$1 million). Thus, it is possible to improve profits and return on marketing communication by appropriately identifying customers for target communications and by matching the channel of communication with customer preferences. The aggregate results suggest that given the improvement of approximately $20 million among a sample of 216 customers, there is a potential for the firm to increase its revenue by at least $1 billion across its entire customer base. However, such benefits may not be realized immediately because the firm also needs to incur costs to move toward a customer-centric view and to train its employees to manage customers on the basis of CLV.

The objective of our study was to analyze the usefulness of CLV as a metric for customer selection and resource allocation strategy. First, we developed and estimated an individual customer-level objective function, the goal of which is to measure CLV. Second, we demonstrated the superiority of selecting customers for contact on the basis of CLV compared with commonly used metrics such as PCR, PCV, and CLD. Third, we evaluated the benefits of designing marketing communications that maximize CLV. We now discuss the implications of our study and how managers can use this knowledge to design efficient marketing programs. We also provide an outline for future researchers to build on the framework proposed herein.

Implications

Antecedents of purchase frequency and contribution margin. The theoretical implications of the purchase frequency model are also related to the CUSAMS customer asset management framework (Bolton, Lemon, and Verhoef 2004). In this study, we tested parts of the CUSAMS framework and found empirical support for the parts we tested. Specifically, the CUSAMS framework proposes that marketing instruments (e.g., direct mailings, reward programs) affect a customer's price perceptions, satisfaction, and commitment. In turn, these affect the length, depth, and breadth of a relationship, which then ultimately influence CLV. We find empirical support for marketing instruments effects on purchase frequency (rich and standardized modes of communication and relationship benefits) and contribution margin (total marketing efforts), both of which ultimately influence CLV. We also find that breadth of purchases (cross-buying) and depth of buying (upgrading) affect purchase frequency, which ultimately influences CLV. In addition, we find support for a nonlinear relationship between supplier communications and purchase frequency. This supports Fournier, Dobscha, and Mick's (1997) expectations that too much communication between suppliers and customers can be disruptive. Thus, our results indicate that managers need to be cautious when designing marketing communication strategies across different channels and need to be wary of contacting customers too many times, especially through rich modes of communication.

We find an inverted U-shaped relationship between returns and purchase frequency. A possible benefit from customers who return products could be the opportunity to understand the reasons for dissatisfaction. In addition, customers who return products within a certain period may do so because they have inherent trust in the supplier and because they expect future benefits, such as improvements in the quality of the product. However, a customer's returning too many times may indicate erosion of trust with the firm or a lower level of future activity. We also find that customers that establish contact through the online channel of communication exhibit high frequency of purchases and have high involvement. Therefore, the online channel can provide an ideal setting for B2B firms to enhance their customer relationships. We find that in addition to influencing purchase frequency, marketing communications influence the expected contribution margin from a customer. Also in the B2B scenario, industry category and size seem to be important factors that influence the magnitude of contribution margin.

Enhancement of marketing productivity. Rust, Zeithaml, and Lemon (2004) propose that firm strategies and tactical marketing actions affect the marketing productivity chain. Our analyses of customer selection investigate how firms can enhance strategies, and our analysis of optimal resource allocation investigates how firms can improve tactical marketing actions.

Customer selection. A CLV metric better identifies customers that provide higher future profits than do PCR, PCV, and CLD. Our analyses indicate that CLV is preferred to incorporate the dynamics of customer purchase behavior into the customer selection process. Managers can substantially improve their return on marketing investments by using a dynamic, customer-level measure of CLV for scoring rather than using the other metrics suggested in the literature and by prioritizing contact programs.

Resource allocation strategy. The results from our study highlight the importance of firms considering individual customers responsiveness to marketing communication as well as the costs involved across various channels of communication when making resource allocation decisions. Our analyses suggest that there is a potential for substantial improvement in CLV through appropriate design of marketing contacts across various channels. When firms design resource allocation rules, they can realize the increase in profits by incorporating the differences in individual customer responsiveness to various channels of communication and the potential value provided by the customer. The proposed resource allocation strategy can be a basis for evaluating the potential benefits of CRM implementations in organizations, and it provides accountability for strategies geared toward managing customer assets.

To summarize, the major conclusions that we derive from our study are the following:

• Marketing communication across various channels affects CLV nonlinearly;

• CLV performs better than other commonly used customer-based metrics for customer selection such as PCR, PCV, and CLD; and

• Managers can improve profits by designing marketing communications that maximize CLV.

Limitations and Further Research

The study has limitations that further studies can address. The results of this study are from a customer database in the high-technology industry. Further studies need to investigate whether the results are generalizable to other industries and settings. In addition, further research needs to develop models that combine forecasts of aggregate competitive responses to marketing actions and customer brand switching with individual-level models of direct marketing. We also consider only the average levels of optimal communication strategy in a channel. However, firms can further improve the efficiency of communication strategy by appropriately sequencing their customer contacts across different channels. In addition, in our study, we provide a framework for maximizing CLV with marketing communications. However, note that we do not compare our proposed resource allocation strategy (that focuses on maximizing CLV) with a strategy that focuses on allocating resources to customers for which the increment in CLV from appropriate design of marketing communication is highest. For example, with an appropriately designed marketing strategy, it is possible that customers that previously had high CLV continue to have high CLV in the future, irrespective of the level of marketing communications, and that some customers that have had low CLV transform to high-CLV customers. Further research can investigate whether the customers selected for high levels of marketing communications are the same when the resource allocation strategy focuses on maximizing CLV or on maximizing incremental CLV. Finally, the sum of optimal CLVs for each individual customer need not lead to the optimal customer equity in the case of a budget constraint. We performed our optimization with a budget constraint, and there were no changes in the substantive results of the study. In addition, our optimization algorithm is flexible enough to allow for inclusion of the budget constraint without any substantial adjustments.

The customer- and supplier-specific antecedents used in the customer response model also can directly affect costs and thus margins. However, because we estimated the customer response model in a single step, which we then included in a net present value function (Equation 1) that includes both costs and margins, we assessed the indirect effect of the covariates on both costs and margins. Further studies can develop and test hypotheses that directly relate CRM efforts to costs and margins. In addition, it can be expected that margins change over time. In this case, the value that a customer provides to a firm is a function of both the expected time frame until the next purchase and the contribution margin at that particular period. We treated several antecedents used in our framework (e.g., upgrading, cross-buying, bidirectional communication, number of returns, number of Web-based contacts) as exogenous variables in our analysis. We used the lagged variables of these to account for potential endogeneity. Further research can investigate more sophisticated techniques that explicitly treat these variables as endogenous. Finally, a notable issue that arises from our analyses is whether the recommendations from an optimization framework pan out when implemented in the real market. Although our study is a step in the right direction to assess the accountability of marketing actions, a field experiment that tests the recommendations of such a framework on a test group against a control group that is managed according to existing norms would provide a stronger justification for CRM-based efforts.

This research study was supported partially by a grant from the Teradata Center for Customer Relationship Management at Duke University, the Marketing Science Institute, and the Institute for Study of Business Markets, and the authors owe special thanks to them. The study greatly benefited from audience discussions at the National Center for Database Marketing conference in Philadelphia and the Marketing Science Institute conference at INSEAD, as well as at Michigan State University, Curtin University, Tilburg University, State University of New York-Buffalo, and the 2003 Marketing Science Conference. The authors thank Munshi Mahfuddin and Rajendra Ladda for research assistance. Special thanks are due to Don Lehmann, Rick Staelin, Greg Allenby, and John Lynch for their comments on the proposal version of this study. The authors thank a business-to-business firm for providing the data for this study. They also thank the anonymous JM reviewers for providing suggestions to enhance the contribution of this study. They thank Renu and Andrew Peterson for copyediting the manuscript.

( n1) In this study, we do not use a budget constraint on the total resources available for contacting customers. Therefore, we are interested only in allocating resources across channels for each individual customer (i.e., within each customer across channels). However, our framework can be applied to allocate resources across channels with each individual customer and across customers in the presence of a budget constraint.

( n2) We also used lagged interpurchase time instead of the log of the lagged interpurchase time, and we did not find any difference in the substantive conclusions. We used log of the lagged interpurchase times because lagged interpurchase times can have a threshold effect on the influence of current interpurchase times (Allenby, Leone, and Jen 1999). The log of the lagged interpurchase time achieved this objective in scaling the tail of the lagged interpurchase time distribution.

( n3) The parameter estimates of the purchase frequency model are based on mean values from 50 repeats with random starting values for each repetition. We adopted such a procedure to ensure that the parameter estimates are global optimal values and are not affected by any local maxima.

( n4) We also estimated the contribution margin model at the monthly, quarterly, and semiannual levels, but we did not find any significant differences in model performance. Thus, to maintain simplicity, we used the contribution margin model with the annual data.

( n5) We also compared the metrics with a censoring time at 18 months and prediction window of 30 months. The substantive results of the study hold even for a 30-month prediction window. The results are available on request from the authors.

( n6) In Equation 2, the level of contacts in each channel for each customer is varied each year. However, to simplify our optimization routine, we assumed that the level of contacts is equal across the prediction period. Our assumption can be viewed as taking the average level of contacts in the prediction periods.

( n7) When forecasting contribution margin for three years ahead, note that we have data on marketing activities from 1997 to 2001. Our last prediction for the long-term analysis is in period 2003, for which we need marketing variables in 2002, which we do not have. For forecasting the contribution margin in 2003, we used the average of marketing-mix variables from 1997 to 2001 for imputing the values in 2002. However, for the optimization phase, the optimized marketing-mix variables are substituted for the 2002 values. We thank a reviewer for bringing this to our attention.

Legend for Chart:

A - Type of Model
B - Representative Research
C - Return on Investment Modeled and Calculated
D - CLV Calculation
E - How Marketing Communication Affects CLV
F - CLV-Based Resource Allocation
G - Resource Allocation for Each Customer
H - Resource Allocation Across Channels
I - Comparison of Customer-Based Metrics
J - Statistical Details

   A               B              C      D      E      F
                                  G      H      I      J

CLV           Berger and
              Nasr (1998)         No    Yes    Yes     No
                                  No     No     No    Yes

              Berger et al.
              (2002)             Yes    Yes    Yes    Yes
                                  No    Yes     No     No

Customer      Blattberg and
equity        Deighton
              (1996)             Yes    Yes     No    Yes
                                  No     No     No    Yes

              Libai,
              Narayandas,
              and Humby
              (2002)             Yes    Yes    Yes     No
                                  No     No     No     No

Database      Reinartz and
marketing     Kumar (2000)       Yes    Yes     No     No
                                  No     No     No    Yes

              Bolton, Lemon,
              and Verhoef
              (2004)             Yes    Yes    Yes     No
                                  No     No     No    Yes

CLV           Reinartz and
antecedents   Kumar (2003)       Yes    Yes    Yes     No
                                  No     No    Yes    Yes

              Rust, Zeithaml,
              and Lemon
              (2004)             Yes    Yes    Yes     No
                                  No     No    Yes    Yes

CLV-based     Berger and
resource      Bechwati
allocation    (2001)             Yes    Yes    Yes    Yes
                                  No     No     No     No

              Present Study      Yes    Yes    Yes    Yes
                                 Yes    Yes    Yes    Yes

TABLE 2
Antecedents, Covariates, and Expected Effects

Legend for Chart:

A - Variable
B - Operationalization
C - Expected Effect
D - Rationale

      A

                             B                            C

                             D

Purchase Frequency Model

Antecedents

Upgrading

             Number of product purchase upgrades          +
             until an observed purchase

             Customers who upgrade have higher
             switching costs with each upgrade,
             which can lead to lower propensity to
             leave and higher recurrent needs
             (Bolton, Lemon, and Verhoef 2004).

Cross-buying

             Number of different product categories       +
             a customer has purchased

             Customers who purchase across
             several product categories have higher
             switching costs and recurrent needs
             (Bowman and Narayandas 2001;
             Reinartz and Kumar 2003).

Bidirectional
communication

             Ratio of number of customer-initiated        +
             contacts to total number of customer
             contacts (customer initiated and
             supplier initiated) between two observed
             purchases

             Two-way communication between
             parties strengthens the relationship and
             ensures that the focal firm is recalled
             when a need arises (Morgan and Hunt
             1994).

Returns

             Number of products the customer              ∩
             returns between two observed
             purchases

             Returns provide an opportunity for firms
             to satisfy customers and ensure repeat
             purchases (Reinartz and Kumar 2003),
             but too many purchases can be
             detrimental to the relationship and can
             indicate that the firm has not used the
             return opportunities appropriately.

Frequency of
Web-based
contacts

             Number of times in a month the               +
             customer contacts the supplier through
             the Internet between two observed
             purchases

             Customers who use online
             communication want transaction
             efficiencies, and customers who want to
             create efficiencies are highly relational
             and have recurring needs (Grewal,
             Corner, and Mehta 2001; Rindfleisch
             and Heide 1997).

Relationship
benefits

             Indicator variable of whether a customer     +
             is a premium service member (based
             on revenue contribution in the previous
             year)

             Acknowledgment of customers with
             relationship benefits reduces the
             propensity of customers to quit and
             increases the probability that the focal
             firm is recalled when a need arises
             (Morgan and Hunt 1994).

Frequency of
rich modes
of communication

             Number of customer contacts by the           ∩
             supplier in a month (through sales
             personnel) between two observed
             purchases

             Timely communication between parties
             reduces the propensity of a customer to
             quit a relationship (Mohr and Nevin
             1990; Morgan and Hunt 1994), but too
             much communication can be detrimental
             to the relationship (Fournier, Dobscha,
             and Mick 1997; Nash 1993); thus, there
             is an optimal communication level.

Frequency of
standardized
modes of
communication

             Number of customer contacts by the           ∩
             supplier in a month (through telephone
             or direct mail) between two observed
             purchases

             Timely communication between parties
             reduces the propensity of a customer to
             quit a relationship (Mohr and Nevin
             1990; Morgan and Hunt 1994), but too
             much communication can be detrimental
             to the relationship (Fournier, Dobscha,
             and Mick 1997; Nash 1993); thus, there
             is an optimal communication level.

Intercontact
time

             Average time between two customer            ∩
             contacts by the supplier across all
             channels of communication between
             two observed purchases

             Along time between contacts can lead
             to forgetfulness, but contacts that are
             too soon can cause dysfunction.

Covariate

Product
category

             Two indicator variables: one indicates a
             hardware purchase; the other indicates
             a software purchase

             A customer's purchase patterns may
             depend on the product category
             purchased.

Contribution Margin Model

Antecedents

Lagged
contribution
margin

             Customer's contribution margin from the      +
             previous year

             Previous revenue is a good predictor of
             current revenue and accounts for any
             model misspecification (Niraj, Gupta,
             and Narasimhan 2001).

Total marketing
efforts

             Total number of customer contacts            +
             across all channels

             Previous marketing communications
             and depth (quantity) of purchases
             positively affect contribution margin
             (Gupta 1988; Tellis and Zufryden 1995).

Total quantity
purchased

             Total quantity of products the customer      +
             purchased across all product categories


             Previous marketing communications
             and depth (quantity) of purchases
             positively affect contribution margin
             (Gupta 1988; Tellis and Zufryden 1995).

Covariates

Size of firm

             Number of employees in the customer          +
             firm

             Control variables that accommodate for
             customer heterogeneity (Niraj, Gupta,
             and Narasimhan 2001).

Industry
category

             Standard industrial classification-based
             industry category to which the customer
             firm belongs

Notes: For the purchase frequency model, the dependent
variable is purchase frequency; for the contribution
margin model, the dependent variable is contribution margin.

Legend for Chart:

A - Variable
B - Mean
C - Standard Deviation
D - Purchase Frequency
E - Upgrading
F - Cross- Buying
G - Bidirectional Communication
H - Returns
I - Frequency of Web-Based Contacts
J - Relationship Benefit-Premium Service Level
K - Frequency of Rich Modes
L - Frequency of Standardized Modes
M - Intercontact Time (Days)

    A                  B          C              D          E

                       F          G              H          I

                       J          K              L          M

Purchase Frequency Model

Purchase             5           8.4          1
frequency           (4.5)       (6.8)

Upgrading            1.32        .89          .62(***)    1
                    (1.16)      (.91)

Cross-buying         2.58        1.7          .53(***)    .39(*)
                    (3.13)      (1.5)

                     1

Bidirectional         .84        2.41         .68(***)    .51(*)
communication        (.62)      (3.52)

                      .48(*)      1

Returns               .91        3.7          .36(***)    .09
                     (.86)      (2.81)

                      .13         .11(*)      1

Frequency of         3.88       25.81         .40(***)    .21
Web-based contacts  (4.37)     (24.94)

                      .15         .17(**)     .06         1

Relationship         .09         .29          .41(***)    .31(**)
benefit-premium     (.12)       (.25)
service
level

                      .36(**)     .22        -.21(*)      .25

                      1

Frequency of         1.79        5.69         .45(***)    .15(**)
rich modes          (1.52)      (5.84)

                      .29(**)     .32(*)      .05         .34(*)

                      .07         1

Frequency of        20.74       47.75        .44(***)    .22(*)
standardized       (22.81)     (45.81)
modes

                      .34(*)      .24(**)     .07         .32(*)

                      .19(*)      .30(**)     1

Intercontact        15.3        13.8          .51(***)   -.08
time               (16.1)      (14.2)
(days)

                      .06        -.01        -.01        -.04

                     -.03        -.11         .06         1

Legend for Chart:

A - Variable
B - Mean
C - Standard Deviation
D - Growth in Contribution Margin
E - Lagged Contribution Margin
F - Total Marketing Efforts
G - Total Quantity Purchase

         A                          B          C           D

                                    E          F           G

Contribution Margin Model

Growth in contribution margin       4955     417,270      1
                                   (4827)   (381,297)

Lagged contribution margin        31,143     323,139   -.78(***)
                                 (32,825)   (318,867)

                                       1

Total marketing efforts             3.49       17.26    .61(***)
                                   (4.5)      (16.24)

                                  .08(*)           1

Total quantity purchased            1.89       13.40    .71(***)
                                   (2.01)     (14.02)

                                 .05(**)         .07      1

(*) Significant at α = 10%.
(**) Significant at α = 5%.
(***) Significant at < 1%.

Notes: The frequency of rich, standardized, and Web-based
contacts represents the frequency of the respective mode of
communication between two consecutive purchases. Values in
parentheses for the descriptive statistics represent Cohort 2.
The correlation matrix for Cohort 2 is similar to that of
Cohort 1 and is available on request from the authors. We do
not report the correlation matrix of the covariates because it
does not have any substantive interpretation.

Legend for Chart:

A - Variable
B - Model 1 (Generalized Gamma Without Mixture)
C - Model 2 (Generalized Gamma with Mixture, Without Temporal
    Variation)
D - Model 3 (Generalized Gamma with Mixture and Temporal
    Variation)

     A
                               B                     C
                                                     D

Component 1

α[sub1]
                        3.05 (4.12)(**)          4.1 (4.2)(**)
                                                3.27 (3.59)(**)

ν[sub1]
                       26.18 (25.08)(**)       42.97 (43.82)(**)
                                               47.05 (48.90)(**)

θ[sub1]
                        .007 (.002)(**)          .01 (.08)(**)
                                                 .04 (.02)(**)

γ[sub1]
                         1.2                     1.2 (1.3)
                                                 1.2 (1.5)

Mass point
                                                 .54 (.55)
                                                 .54 (.56)

Component 2

α[sub1]
                                                 .58 (1.57)(**)
                                                1.48 (3.61)(**)

nu;[sub1]
                                               38.21 (34.28)(**)
                                               32.07 (49.02)(**)

θ[sub1]
                                                .008 (.012)(**)
                                                .005 (.001)(**)

γ[sub1]
                                                  .9 (.9)
                                                  .9 (.9)

Mass point
                                                 .46 (.45)
                                                 .46 (.44)

Coefficients

β[sub01]
                                                 .89 (-1.52)(**)
                                               -2.05 (-2.51)(*)

Lagged log interpurchase
time
                                               -2.09 (-2.85)(**)

Antecedents: Customer Characteristics

Upgrading
                                                5.01 (5.62)(**)

Cross-buying
                                                6.08 (6.92)(**)

Bidirectional communication
                                                1.49 (1.01)(**)

Returns
                                               10.52 (9.98)(**)

(Returns)²
                                               -3.84 (-4.01)(*)

Frequency of Web-based
contacts
                                                3.52 (2.38)(**)

Antecedents: Supplier-Specific Factors

Relationship benefits
                                                9.78 (7.65)(**)

Frequency of rich modes of
communication
                                                4.50 (5.65)(**)

(Frequency of rich modes of
communication)²
                                               -1.30 (-1.28)(**)

Frequency of standardized
modes of communication
                                                6.53 (7.02)(**)

(Frequency of standardized
modes of communication)²
                                                -.28 (-.53)(**)

Intercontact time
                                                9.64 (8.56)(**)

(Intercontact time)²
                                               -3.21 (-4.05)(**)

Log-likelihood
                    -3298.91 (-3627.08)     -2908.66 (-3297.04)
                                            -2237.82 (-2437.91)

AIC
                       -6606 (-7262)           -5825 (-6602)
                                               -4484 (-4884)

BIC
                       -6639 (-7295)           -5910 (-6687)
                                               -4683 (-5083)

Relative absolute error
                         .95 (.91)               .84 (.80)
                                                 .51 (.53)

(*) Posterior sample values between the 2.5th and 97.5th
percentile do not contain zero.

(**) Posterior sample values between the .5th percentile and
99.5th percentile do not contain zero.

Notes: Values in parentheses represent Cohort 2. The product
category variable was not significant in our analysis, and
thus we do not include it here. Relative absolute error is
with respect to a moving average model. The significance
levels apply to the coefficients of Cohorts 1 and 2.

Legend for Chart:

A - Independent Variable
B - Parameter Estimate

          A                                          B

Intercept                                  N.S.        (N.S.)
Contribution in t-2                        .83(***)    (.85(***))
Lagged total quantity purchased            .02(**)     (.03(**))
Size                                       .02(**)     (.02(**))
Aerospace                                  N.S.        (N.S.)
Financial services                         .02(*)      (.02(**))
Manufacturing                              N.S.        (N.S.)
Technology                                 .03(**)     (.02(**))
Consumer packaged goods                    .03(***)    (.03(***))
Education, K-12                           -.03(***)   (-.02(***))
Travel                                     N.S.        (N.S.)
Government                                 .02(*)      (.02(*))
Lagged total level of marketing effort     .04(***)    (.06(***))

(*) Significant at α = .10.
(**) Significant at α = .05.
(***) Significant at α < .01.

Notes: The reported coefficients are standardized estimates; the
values in parentheses represent Cohort 2. N.S. = not significant.

Legend for Chart:

A - Percentage of Cohort (Selected from Top)
B - CLV
C - PCR
D - PCV
E - CLD

        A                B         C          D          E

5%

Gross profit ($)      144,883    71,908    131,735    107,719
Variable costs ($)      1,588       979        950        790
Net profit ($)        143,295    70,929    130,785    106,389

10%

Gross profit ($)       78,401    27,981     72,686     55,837
Variable costs ($)      1,245       943        794        610
Net profit ($)         77,156    27,038     71,892     55,227

15%

Gross profit ($)       56,147    15,114     52,591     44,963
Variable costs ($)        807       944        809        738
Net profit ($)         55,340    14,170     51,782     44,225

Notes: All metrics are evaluated at 30 months, with an 18-month
prediction window. Cohort 2 provides similar results. The
reported values are cell medians. Gross profit is residual
revenue after removing cost of goods sold. In general, for the
firm that provided the database, the cost of goods sold is
approximately 70%; thus, gross profit = revenue x .3.

DIAGRAM: FIGURE 1; A Conceptual Framework for Measuring and Using CLV

Allenby, Greg M., Robert P. Leone, and Lichung Jen (1999), "A Dynamic Model of Purchase Timing with Application to Direct Marketing," Journal of American Statistical Association, 94 (June), 365-74.

Balakrishnan, P.V. and Varghese A. Jacob (1996), "Genetic Algorithms of Product Design," Management Science, 42 (8), 1105-1118.

Baltagi, Badi H. (1998), Econometric Analysis of Panel Data. New York: John Wiley & Sons.

Bell, David, John Deighton, Werner J. Reinartz, Roland T. Rust, and Gordon Swartz (2002), "Seven Barriers to Customer Equity Management," Journal of Service Research, 5 (1), 77-85.

Berger, Paul D. and Nada Nasr Bechwati (2001), "The Allocation of Promotion Budget to Maximize Customer Equity," OMEGA, 29 (1), 49-61.

-----, Ruth N. Bolton, Douglas Bowman, Elten Briggs, V. Kumar, A. Parasuraman, and Creed Terry (2002), "Marketing Actions and the Value of Customer Assets: A Framework for Customer Asset Management," Journal of Service Research, 5 (1), 39-54.

------ and Nada I. Nasr (1998), "Customer Lifetime Value: Marketing Models and Applications," Journal of Interactive Marketing, 12 (Winter), 17-30.

Blattberg, Robert C. and John Deighton (1996), "Manage Marketing by the Customer Equity Test," Harvard Business Review, 74 (July August), 136-44.

------, Gary Getz, and Jacquelyn S. Thomas (2001), Customer Equity: Building and Managing Relationships as Assets. Boston: Harvard Business School Press.

----- and Edward Malthouse ((2003), "Evaluating the Predictive Accuracy of Long-Term Value Models for Customer Relationship Management Strategies," working paper, Department of Marketing, Northwestern University.

Bolton, Ruth and Katherine N. Lemon (1999), "A Dynamic Model of Customers Usage of Services: Usage as an Antecedent and Consequence of Satisfaction," Journal of Marketing Research, 36 (May), 171-86.

-----, -----, and Peter C. Verhoef (2004), "The Theoretical Underpinnings of Customer Asset Management: A Framework and Propositions for Future Research," Journal of the Academy of Marketing Science, 32 (3), 271-92.

Bowman, Douglas and Das Narayandas (2001), "Managing Customer-Initiated Contacts with Manufacturers: The Impact on Share of Category Requirements and Word-of-Mouth Behavior," Journal of Marketing Research, 38 (August), 281-97.

Cannon, Joseph P. and Christian Homburg (2001), "Buyer Seller Relationships and Customer Firm Costs," Journal of Marketing, 65 (January), 29-43.

Cowles, Mary K. and Bradley P. Carlin (1996), "Markov Chain Monte Carlo Convergence Diagnostics: A Comparative Review," Journal of American Statistical Association, 91 (June), 883-904.

Del Moral, P. and L. Miclo (2001), "Asymptotic Results for Genetic Algorithms with Applications to Nonlinear Estimation," in Theoretical Aspects of Evolutionary Computation, L. Kallel, B. Naudts, and A. Rogers, eds. Berlin: Springer-Verlag, 439-94.

Dhar, Ravi and Rashi Glazer (2003), "Hedging Customers," Harvard Business Review, 81 (May), 3-8.

Dorsey, Robert E. and Walter J. Mayer (1995), "Genetic Algorithms for Estimation Problems with Multiple Optima," Journal of Business and Economic Statistics, 13 (1), 53-67.

Dwyer, Robert F., Paul H. Schurr, and Sejo Oh (1987), "Developing Buyer Seller Relationships," Journal of Marketing, 51 (April), 11-27.

Fournier, Susan, Susan Dobscha, and David Glen Mick (1997), "Preventing the Premature Death of Relationship Marketing," Harvard Business Review, 75 (January February), 2-8.

Ganesan, Shankar (1994), "Determinants of Long-Term Orientation in Buyer Seller Relationships," Journal of Marketing, 58 (April), 1-19.

Gelman, A. and D.B. Rubin (1992), "Inference from Iterative Simulation Using Multiple Sequences," Statistical Science, 7 (3), 457-511.

Goldberg, David E. (1989), "A Simple Genetic Algorithm," in Genetic Algorithms in Search Optimization and Machine Learning. Boston: Addison-Wesley Longman, 10-14.

Grewal, Rajdeep, James M. Corner, and Raj Mehta (2001), "An Investigation into the Antecedents of Organizational Participation in Business-to-Business Electronic Markets," Journal of Marketing, 65 (July), 17-34.

Gupta, Sunil (1988), "The Impact of Sales Promotion on When, What, and How Much to Buy," Journal of Marketing Research, 25 (November), 342-55.

Hogan, John E., Donald R. Lehmann, Mario Merino, Rajendra K. Srivastava, Jacquelyn S. Thomas, and Peter C. Verhoef (2002), "Linking Customer Assets to Financial Performance," Journal of Service Research, 5 (August), 26-38.

Jap, Sandy D. (2001), "The Strategic Role of Salesforce in Developing Customer Satisfaction Across the Relationship Lifecycle," Journal of Personal Selling & Sales Management, 21 (2), 95-108.

Kamakura, Wagner and Gary Russell (1989), "A Probabilistic Choice Model for Market Segmentation and Elasticity Structure," Journal of Marketing Research, 26 (November), 379-90.

Liang, Faming and Wing Hung Wong (2001), "Real-Parameter Evolutionary Monte Carlo with Application to Bayesian Mixture Models," Journal of the American Statistical Association, 96 (454), 653-66.

Libai, Barak, Das Narayandas, and Clive Humby (2002), "Toward an Individual Customer Profitability Model: A Segment Based Approach," Journal of Service Research, 5 (1), 69-76.

Midley, David F., Robert E. Marks, and Lee G. Cooper (1997), "Breeding Competitive Strategies," Management Science, 43 (3), 257-75.

Mohr, Jakki and John R. Nevin (1990), "Communication Strategies in Marketing Channels: A Theoretical Perspective," Journal of Marketing, 54 (October), 36-51.

Morgan, Robert M. and Shelby D. Hunt (1994), "The Commitment Trust Theory of Relationship Marketing," Journal of Marketing, 58 (July), 20-38.

Naik, Prasad A., Murali K. Mantrala, and Alan G. Sawyer (1998), "Planning Media Schedules in the Presence of Dynamic Advertising Quality," Marketing Science, 17 (3), 214-35.

Nash, Edward L. (1993), Database Marketing: The Ultimate Marketing Tool. New York: McGraw-Hill.

Neal, Radford M. (2000), "Slice-Sampling," Technical Report No. 2005, Department of Statistics, University of Toronto.

Newton, M.A. and A.E. Raferty (1994), "Approximate Bayesian Inference with the Weighted Likelihood Bootstrap," Journal of the Royal Statistical Society, (Series B, 56), 3-48.

Niraj, Rakesh, Mahendra Gupta, and Chakravarthi Narasimhan (2001), "Customer Profitability in a Supply Chain," Journal of Marketing, 65 (July), 1-16.

Reinartz, Werner J. and V. Kumar (2000), "On the Profitability of Long-Life Customers in a Noncontractual Setting: An Empirical Investigation and Implications for Marketing," Journal of Marketing, 64 (October), 17-35.

----- and ----- ((2002), "The Mismanagement of Customer Loyalty," Harvard Business Review, 80 (7), 86-94.

----- and ----- ((2003), "The Impact of Customer Relationship Characteristics on Profitable Lifetime Duration," Journal of Marketing, 67 (January), 77-99.

Rindfleisch, Aric and Jan B. Heide (1997), "Transaction Cost Analysis: Past, Present, and Future Applications," Journal of Marketing, 61 (October), 30-55.

Roberts, Mary Lou and Paul Berger (1999), Direct Marketing Management, 2d ed. Upper Saddle River, NJ: Prentice Hall.

Rust, Roland T., Valarie A. Zeithaml, and Katherine N. Lemon (2004), "Return on Marketing: Using Customer Equity to Focus Marketing Strategy," Journal of Marketing, 68 (January), 23-53.

Sawhney, Mohanbir (2001), "Don't Homogenize, Synchronize," Harvard Business Review, 81 (July August), 101 108.

Schmittlein, David and Donald G. Morrison (1985), "Is the Customer Alive? The American Statistician," 39 (4), 291-95.

----- and Robert A.. Peterson (1994), "Customer Base Analysis: An Industrial Purchase Process Application," Marketing Science, 13 (1), 41-67.

Shepard, David (2001), The New Direct Marketing: How to Build a Profit-Driven Database Marketing Strategy, 3d ed. New York: McGraw-Hill.

Sheth, Jagdish N. and Atul Parvatiyar (1995), "Relationship in Consumer Markets: Antecedents and Consequences," Journal of the Academy of Marketing Science, 23 (4), 255-71.

Tellis, Gerard and Fred Zufryden (1995), "Tackling the Retailer Decision Maze: Which Brands to Discount, How Much, When, and Why?" Marketing Science, 14 (3), 271-300.

Thompson, Bob (2001), "CRM and the Internet: Leading Edge Strategies for a Multi-Channel World," ((accessed May 19, 2004), [available at http://www.businessweek.com/adsections/ care/crm/crm_index.htm].

Venkatesan, Rajkumar, Trichy V. Krishnan, and V. Kumar (2004), "Evolutionary Estimation of Macro-Level Diffusion Models," Marketing Science, forthcoming.

Villas-Boas, Miguel J. and Russell S. Winer (1999), "Endogeneity in Brand Choice Models," Management Science, 45 (10), 1324-38.

Purchase Frequency Model

Model estimation. For Cohorts 1 and 2, the results are based on 50,000 samples of the MCMC algorithm (for details of the algorithm, see Allenby, Leone, and Jen 1999, Appendix). We simulated the posterior distribution using five parallel chains with overdispersed starting values. In each chain, we used the initial 40,000 iterations as burn-in and used the last 10,000 iterations to obtain posterior statistics. We used "slice-sampling" ((Neal 2000) to obtain random samples. We assessed the autocorrelations of the posterior samples to perform thinning. The autocorrelation functions revealed that every fifth sample is unrelated to every other fifth sample. Thus, the posterior statistics are based on 2000 samples (using every fifth sample from 10,000 samples). We assessed the convergence of the algorithm from the line plots of the posterior sample and by evaluating Gelman and Rubin's (1992) √R statistic. Values of √R that are closer to 1 reveal that all the chains have converged to true posterior distribution. In our analyses, we found that the value of √R ranged from 1.2 to .9 (in previous research [Cowles and Carlin 1996], these values have been found to indicate convergence of the MCMC chains). Our approach to investigating within-chain autocorrelations, computing Gelman and Rubin's statistic, and visually inspecting the sampling plots has been recommended widely for convergence diagnostics (Cowles and Carlin 1996). We implemented the estimation algorithm using the GAUSS software package, and we performed the convergence diagnostics using CODA. We also used multiple prior values for the estimates and did not find a significant impact on the posterior samples; this is because we used diffuse prior values and because of the large sample size of our data set.

Model selection. Table 4 shows that a simple generalized gamma model without the probit link function or time-varying covariates (Model 1) provides a log-likelihood of -3298.91 (we calculated the log-likelihood using Newton and Raferty's [1994] log marginal density measure)). A generalized gamma model with the probit link function (two subgroups) but no time-varying covariates (Model 2) provides a log-likelihood of 2908.66. Finally, the loglikelihood for the generalized gamma model with the probit link function (with two subgroups) and time-varying covariates (Model 3) is 2237.82. Increasing the number of subgroups in Models 2 or 3 did not result in significantly higher log-likelihoods. We also computed the AIC and BIC for model selection. For both measures, a higher value indicates a better model. We find that Model 3 has the highest value for both AIC and BIC (AIC = -4484 for Model 3, -5825 for Model 2, and -6606 for Model 1; BIC = -4683 for Model 3, 5910 for Model 2, and -6639 for Model 1). In addition, a latent-class finite mixture model with two segments provides a likelihood of -2938. Overall, the results show that the generalized gamma model with two subgroups provides a good fit to the data.

Distribution parameters. The values of γ in the generalized gamma model are γ[sub1] = 1.2 and γ[sub2] = .9 for Subgroups 1 and 2, respectively. The component masses for Subgroup 1 (φ[sub1]) and Subgroup 2 (φ[sub2]) are .54 and .46, respectively. All the distribution parameters (φ[subk], v[subk] and Θ[subk] have more than 99% of their samples different from zero. The mean expected interpurchase time for Subgroup 1 is 4.2 purchases in a year, and the mean expected frequency for Subgroup 2 is 1.01 purchases in a year. Given the variation in expected frequencies in each subgroup, we term Subgroup 1 the "active state" and Subgroup 2 the "inactive state".

Influence of antecedents and covariates. Our empirical analyses support all the proposed effects of the antecedents on purchase frequency.

Out-of-sample forecasting accuracy. We used relative absolute error (RAE) to evaluate the forecasting accuracy of the generalized gamma model compared with that of a naive moving average model (the moving average model is implemented as an updated average of every consecutive interpurchase time for a customer). We used all but the last observation for each customer (calibration sample) to estimate the parameters of each model. We then used the mean posterior values to compute the expected time until next purchase from Equation 3. We then compared the forecast time until the next purchase with the holdout sample (formed from the last observation for each customer) to assess the mean absolute deviation (MAD). The RAE is given by the ratio of the model MAD to the MAD based on the moving average measure. Based on the RAE measure, the generalized gamma model with time-varying covariates (Model 3) has an RAE of .51, compared with that of a naive moving average technique. The MAD from Model 3 is 5.21 months, compared with 10.24 months for the moving average measure. Model 3 provides the best improvement in forecasting accuracy compared with Model 2 (RAE = .84) and Model 1(RAE = .95). We also assessed the predictive capability of the purchase frequency model using hit rate. In other words, we assessed the number of purchases in the holdout sample that the model also correctly predicted as a purchase. We observe that among customers who bought within 12 months, the model currently identifies 89% of them, and among customers who did not buy within 12 months, the model currently identifies 90% of them.

Contribution Margin Model

Estimation and influence of antecedents and covariates. We estimated the contribution margin model on annual data from 1997 (t - 4)) to 2000 (t) for Cohort 1 and from 1998 to 2000 for Cohort 2. The revenue in 2001 (t + 1) is a holdout sample for Cohorts 1 and 2. The coefficients of the contribution margin model are provided in Table 5. The contribution margin model provides an adjusted R2 of .68. Overall, the results of the analyses support all the hypothesized relationships.

Out-of-sample forecasting accuracy. As with typical time-series models, we advanced the independent variables by one period to forecast the dependent variable in period t+ 1. Specifically, when the contribution margin model is used to predict period t + 3, the contribution margin in period t + 1 is an independent variable and is obtained from the prediction in period t + 1. The contribution margin model predicts the growth in revenue from period t to t + 1. We obtained the magnitude of revenue in period t + 1 by adding the predicted value from the contribution margin model to the base revenue in period t. We evaluated the performance of the contribution margin model in the holdout sample on the basis of our estimates from the calibration sample. Table A1 provides the descriptive statistics of the observed contribution margin and the predicted contribution margin in the holdout sample (period t + 1). In the holdout sample, the mean predicted contribution margin in period t+ 1 is approximately $67,729, and the mean observed contribution margin is approximately $64,396.

CLV

We entered predictions from the purchase frequency (Equation 3) and contribution margin (Equation 4) models into Equation 2 to obtain the net present value of future profits (period t + 1) from each customer. The purchase frequency model predicts the expected time in months until next purchase for each customer. We assigned a 30% margin after accounting for cost of goods sold (the managers who provided the database informed us that 30% was a nominal margin for most of their products), and we computed the variable costs using costs of communication. The mean unit cost of standardized modes of communication is approximately $3, and the mean unit cost of rich modes of communication is $60. We computed the unit cost of communication for each customer as the ratio of the total contacts for a given channel in a given year to the total cost of contact for a given mode in a given year. Finally, we used an annual discount rate of 15% for each customer, which is based on the lending rate that is appropriate for the time of the study.

PCR and PCV

We define PCR as the revenue provided by the customer in the most recent observed purchase. We define PCV as the cumulative profits obtained from a customer until the current period. The cumulative profits are calculated annually from a customer's initiation until the current period. We projected the profit in each year to current terms using a discount factor. The PCV calculation is as follows:

(B1) [Multiple line equation(s) cannot be represented in ASCII text]

where CM[subi,t] is the contribution margin for customer i in period t; MC[subi,t] denotes marketing costs for customer i in period t; t is an index for time period (t = 0 for the period of customer initiation; for example, t = 0 for 1997 for Cohort 1 customers, and t = 0 for 1998 for Cohort 2 customers); T is the current period; and r is the discount rate, which we set at 15%.

CLD

In our analysis, we evaluated the probability that a customer is alive or dead in the planning window using the P(Alive) measure that Schmittlein and Peterson (1987) and Reinartz and Kumar (2002) recommend. The P(Alive) measure uses the previous purchase pattern to predict the probability that a customer is still alive at each period in the prediction window. Higher values of P(Alive) indicate longer lifetime duration.

Researchers in marketing have only recently begun to recognize the potential benefits of using genetic algorithms (Balakrishnan and Jacob 1996; Midley, Marks, and Cooper 1997; Naik, Mantrala, and Sawyer 1998; Venkatesan, Krishnan, and Kumar 2004) in deriving optimal strategies for complex marketing problems. The genetic algorithm proceeds by searching for the optimal level of contact for each customer that maximizes CLV. The sum of the optimal CLVs from each individual customer provides the optimal customer equity of the analysis sample.( n7) Specifically, we varied the contact levels for each customer and then calculated the sum of CLV of all customers in the sample. Our objective was to calculate the maximum value for this sum of the CLVs. In this case, our optimization algorithm maximized the objective function by varying 432 parameters in Cohort 1 (216 customers and 2 parameters for each customer [levels of rich and standardized modes]). Following research in customer equity (Rust, Zeithaml, and Lemon 2004), we set the time frame for our optimization framework as three years. Our database provides information on the approximate unit cost of communication through rich modes and standardized modes to each customer. On average, the unit marketing cost through standardized modes (average of direct mail and telephone sales) is $3, and the unit marketing cost through rich modes (salesperson contacts) is $60. Thus, the cost of communication through rich modes is approximately 20 times the cost of communication through standardized modes. Such a cost index is commonly encountered. We set the parameters in the genetic algorithm as follows: population size = 200, probability of crossover = .8, probability of mutation = .25, and convergence criteria = difference in solution in the last 10,000 iterations should be less than .01%. We ran the genetic algorithm at least 50 times and used the mean of the resource levels corresponding to the maximum CLV from each run as the resource reallocation rule for each customer.

Comparison of Descriptive Statistics Between Observed Contribution Margin and Predicted Contribution Margin in the Holdout Sample

Legend for Chart:

B - Mean
C - Standard Deviation
D - Minimum
E - Maximum

    A             B           C          D            E

Observed        50,199      23,850      -171      1,010,881
               (49,229)    (24,836)    (-181)    (1,420,981)

Predicted       57,729      24,538      -178      1,897,257
               (74,283)    (22,598)    (-159)    (1,938,458)

Notes: All reported values are in dollars and are rounded to
the nearest integer. Values in parentheses represent Cohort 2.

~~~~~~~~

By Rajkumar Venkatesan and V. Kumar

Rajkumar Venkatesan is Assistant Professor in Marketing (e-mail: RVenkatesan@sba.uconn.edu), and V. Kumar is ING Chair Professor in Marketing and Executive Director, ING Center for Financial Services (email: vk@sba.uconn.edu), School of Business, University of Connecticut.

A Customer Relationship Management Roadmap: What Is Known, Potential Pitfalls, and Where to Go 

The goal of this preface is to describe how the special section on customer relationship management (CRM) was developed. In May 2003, Richard Staelin, Executive Director of the Teradata Center for Customer Relationship Management at Duke University, proposed that Journal of Marketing (JM) publish a special section. The proposal included activities that were designed to promote interactions among marketing academics and practitioners; the goal was to stimulate dialogue and new research on CRM. I found the proposal attractive because CRM is a broad-based topic that interests many marketers. After extensive discussion, the American Marketing Association (AMA) and the Teradata Center formally agreed to cosponsor the special section. Subsequently, there was a conference on Relationship Marketing and Customer Relationship Management (cochaired by Michael Ehret, Wesley Johnston, Michael Kleinaltenkamp, and Lou Pelton) that took place at Freie Universität Berlin in the summer of 2003;( n1) a conference on Customer Management (cosponsored by the Marketing Science Institute and the Teradata Center) that was held at Duke University in March 2004; and two special sessions on CRM that were featured at the AMA Winter Educators' Conference held in San Antonio, Tex., in February 2005. The conferences provided many opportunities for dialogue, and the response from marketers who attended these events was enthusiastic. I also invited Richard Staelin and William Boulding (Executive Codirector of the Teradata Center) to work with me as consulting editors for the special section, and they agreed. A call for papers requested that authors submit their manuscripts to JM by May 2004. The consulting editors and I evaluated every submission with the assistance of an expert panel that included Leonard Berry, John Deighton, Michael Ehret, Christian Grönroos, Sunil Gupta, Wayne Hoyer, Wagner Kamakura, Wesley Johnston, Donald R. Lehmann, Charlotte Mason, Carl Mela, Scott Neslin, Roland Rust, Michel Wedel, and Valarie Zeithaml. All submissions underwent JM 's standard double-blind review process, and members of JM's editorial review board served as reviewers. I would like to express my appreciation to everyone who participated in the development of the special section. The culmination of our work together is a set of nine articles and two essays that advance the science and practice of CRM. I hope that these articles stimulate new intellectual discoveries.

This article introduces the ten articles that appear in this special section on customer relationship management (CRM). An overarching goal of this article is to provide the reader with a roadmap that places these articles in the context of the CRM landscape. We suggest 11 propositions about what is known about CRM and the potential pitfalls and unknowns that firms face in the implementation of CRM. We also provide six recommendations for further CRM research. We organize our discussion around these themes before offering concluding comments.

Before assessing what is known about CRM, we begin by placing the field of CRM in the overall context of marketing thought.( n2) Many years ago, economists introduced the concept of value maximization, whereby a firm maximizes profits and consumers maximize utility. Today, we have the concept of CRM. Theorists in this area still emphasize firm performance and customer value, though they also talk about the dual creation of firm and customer value (Payne and Frow 2005; Rogers 2005; Vargo and Lusch 2004). The question we raise is whether the field's focus on CRM sheds light on the understanding of customer and firm behavior or whether it just creates more "heat." Many vendors argue that CRM requires a paradigm shift in firm behavior. If this is true, CRM is truly a really big new idea. However, others contend that the concepts of CRM are not fundamentally different from what economists put forward many years ago. If this is so, the following questions arise: Is CRM anything other than a repackaging of basic marketing ideas that have extended and built on the classic economic paradigm? Should CRM be viewed simply as one of many jargon-laden fads that have come and gone in the business world? Or is a third explanation possible? Namely, does CRM represent the evolution and integration of marketing ideas and newly feasible and cost-effective technologies? In this view, CRM is neither a fad nor a paradigm shift. After observing the development of the CRM field, we offer the following proposition:

P₁: CRM is the outcome of the continuing evolution and integration of marketing ideas and newly available data, technologies, and organizational forms.

To support this proposition, we briefly document this evolution.

One of the original big ideas in marketing is that for firms to stay in existence, they should not focus on selling products but rather on fulfilling needs (Levitt 1960). Thus, a drill manufacturer is in the business of providing a customer a hole, and a railroad company is in the business of providing transportation. This is a key component of CRM because the emphasis is not on how to sell the product but rather on creating value for the customer and, in the process, creating value for the firm (staying in existence). In other words, it is a process of dual creation of value. Levitt (1969) introduced the concept of the augmented product, stressing that consumers are interested in the total buying experience, not just the core product. Again, CRM relies on this concept because it tries to find the specific elements of the exchange process that produce value to the customer.

Bagozzi (1974) refocused people's attention on the actual exchange process by reiterating the fundamental economic concept that an exchange occurs only when both parties perceive that they are receiving value. Almost ten years later, Berry (1983) shifted the emphasis to the relationship between the company and the customer. At the time, his interest was in the service sector and the need for the service organization to attract customers and then maintain and enhance these customer relationships. On the basis of his ideas and related conceptual work (Arndt 1979; MacNeil 1978; Morgan and Hunt 1994), the concept of building relationships was expanded to several different domains, such as industrial buyer-seller relationships (Dwyer, Schurr, and Oh 1987) and channels of distribution (Gaski 1984). Others adopted the idea of building relationships and extended it conceptually in various ways (Boulding et al. 1993; Grönroos 1994; Gummesson 1987; Webster 2002). This body of literature discusses concepts that are relevant to CRM, such as the influence of prior experience on future customer expectations, the different treatment of each customer, and the value of long-term relationships.

Concurrently, other marketing scholars turned their attention to the core capabilities of the firm that were necessary to develop and maintain good customer relationships. In some sense, this was a formalization of the concept and processes implied by the "three Cs" (i.e., customer, company, and competitor) analysis. As a result, concepts such as market orientation (Kohli and Jaworski 1990; Narver and Slater 1990), market focus (Day 1994), and market-based learning (Vorhies and Hunt 2005) were developed that emphasized the establishment of good information processes and capabilities within the firm to understand the needs and wants of customers, thus making firms more efficient and effective in managing customer relationships. In addition, there was an evolution from product, or brand, management to customer management (Sheth 2005) and from product portfolio management to customer portfolio management (Johnson and Selnes 2004). These transitions were due in part to work in the area of brand equity, which recognized that equity resides in the minds of consumers (Keller 1993); this shifted the locus of attention from brands and products to customers.

With these developments in marketing as a backdrop, there was an explosion of customer data in the 1980s. Although some attempts were made to organize these data for analytic purposes, many firms were overwhelmed by this onslaught of potentially useful information. In anticipation of hardware and software solutions to these data problems, Peppers and Rogers (1993) introduced the concept of one-to-one marketing, and Pine (1993) introduced the concept of mass customization. Vendors capitalized on these ideas with hardware and software solutions and began using the term CRM to refer to the collection of data and activities surrounding the management of the customer-firm interface. These CRM solutions enabled firms to acquire, warehouse, and analyze data about customer behavior and company actions more easily. Using these data and analyses, firms began to focus on acquiring new customers; retaining their current customers (i.e., building long-term relationships); and enhancing these relationships through such activities as customized communications, cross-selling, and the segmentation of customers, depending on their value to the firm (Payne and Frow 2005). Implementation of these CRM solutions also required firms to have a customer relational orientation (Jayachandran et al. 2005; Srinivasan and Moorman 2005) and to have processes in place to collect, analyze, and apply the acquired customer information (Jayachandran et al. 2005).

Thus, the question is, What is new about CRM? On the basis of our preceding discussion, it could be argued that CRM is the relabeling of a mixture of different marketing ideas in the extant marketing literature. However, we believe that CRM represents an evolution beyond a repackaging of existing ideas. Specifically, we posit that CRM goes beyond extant literature because it "requires a cross-functional integration of processes, people, operations, and marketing capabilities that is enabled through information, technology, and applications" (Payne and Frow 2005, p. 168). Indeed, CRM goes beyond a customer focus. Not only does CRM build relationships and use systems to collect and analyze data, but it also includes the integration of all these activities across the firm, linking these activities to both firm and customer value, extending this integration along the value chain, and developing the capability of integrating these activities across the network of firms that collaborate to generate customer value, while creating shareholder value for the firm.

P₂: The field of CRM has begun to converge on a common definition.

Payne and Frow (2005) document numerous definitions of CRM in the literature (see their Appendix). These definitions range from CRM as the implementation of specific technology solutions to a holistic approach of managing customer relationships that simultaneously creates both customer and firm value. This plethora of definitions has caused some confusion. Parvatiyar and Sheth (2001) note that a prerequisite for an emerging field to coalesce into an established field is for the discipline to establish an acceptable definition that captures all the major aspects of the concept. Payne and Frow attempt to provide such a definition. It is possible to quibble about the specific wording, but we agree with the basic elements of their definition. Specifically, CRM relates to strategy, the management of the dual creation of value, the intelligent use of data and technology, the acquisition of customer knowledge and the diffusion of this knowledge to the appropriate stakeholders, the development of appropriate (long-term) relationships with specific customers and/or customer groups, and the integration of processes across the many areas of the firm and across the network of firms that collaborate to generate customer value.

In addition to theoretical development, a prerequisite for the applied development of CRM is that it should demonstrably enhance firm performance. This is a necessary quality in the evaluation of any firm or marketing activity (e.g., Lehmann 2004; Rust et al. 2004). With this in mind, note that it is not necessarily a widely held belief that the implementation of CRM activities leads to firm value. To this end, consider the numerous articles that appear in the business press (e.g., Rigby, Reichheld, Schefter 2002; Whiting 2001). Nonetheless, we propose the following:

P₃: Companies have developed proven CRM practices that enhance firm performance.

Eight of the ten articles in this special section directly address this proposition. These articles use different measures of performance in many different contexts, and they use various research methods. However, all eight articles demonstrate that CRM activities can enhance firm performance. For a field that has come under attack for not meeting this objective, we believe that this is a powerful result.

Using a case study approach, Ryals (2005) shows that one of the business units she studied was able to achieve a 270% increase in business unit profits (above target) by implementing several straightforward CRM measures. Using a multifirm (cross-sectional) database, Srinivasan and Moorman (2005) show that firms that invest more in CRM activities and technology have greater customer satisfaction. Using another multifirm database, Mithas, Krishnan, and Fornell (2005) show that the use of CRM applications is associated with increased customer knowledge, which in turn is associated with greater customer satisfaction. Using yet another multifirm database, Jayachandran and colleagues (2005) show that firm performance measured in terms of retention and customer satisfaction is greater for firms that have good relational information processes in place.

Cao and Gruca (2005), Lewis (2005), Thomas and Sullivan (2005), and Gustafsson, Johnson, and Roos (2005) all use data collected within a single firm over time. Cao and Gruca, Lewis, and Thomas and Sullivan use data from both the firm and its customers to develop specific CRM applications to increase the firm's performance. Cao and Gruca center their attention on acquiring the "right" customers; Lewis provides a process that identifies and considers dynamic customer behavior, thus enabling a pricing scheme that increases long-term profits; and Thomas and Sullivan develop a decision support system using an enterprise database that allows the firm to modify its communication message depending on where particular customers live and how they shop. In each case, the authors show how firm profits can be increased. Gustafsson, Johnson, and Roos (2005) examine customer behavior over time and show that some of the intermediate relationship performance measures that emerge from the business-to-business literature (e.g., satisfaction, calculative commitment) directly and positively influence actual behavior in the form of retention within a business-to-consumer setting.

We must emphasize four points here. First, the eight empirical articles in this special section demonstrate the positive impact of CRM in a wide variety of industry settings. Thus, success with CRM is not contingent on being a part of a particular industry (e.g., financial services). Second, we note that all of the application articles are narrow rather than comprehensive. Thus, they find local improvements in profits. We can only speculate about what could be accomplished with a more comprehensive systems approach; we also express some concern that local solutions can sometimes be suboptimal in the long run. Third, only Ryals's (2005) study directly measures both the costs and the revenues associated with the CRM activities to assess overall profits. Lewis (2005) and Cao and Gruca (2005) examine profits, but because of data unavailability, they must make assumptions about costs to generate these numbers. All the other studies use proxies for profits. Because several of the studies use customer satisfaction for their performance measure, it is of interest that Gustafsson, Johnson, and Roos (2005) show that customer satisfaction is negatively associated with observed customer churn, thus providing strong evidence that customer satisfaction is a useful precursor of downstream outcomes. However, it is important to note that this same research indicates that satisfaction is not the only predictor of downstream performance measures.

This leads to a fourth observation about CRM activities and firm performance. Payne and Frow (2005) emphasize that one major element in any CRM system is the measurement process. Although the ultimate objective of any measurement process is to increase shareholder value, one of the real advantages of a CRM measurement process is that the firm normally also obtains measures such as customer lifetime value and acquisition and retention costs, which relate to the value dual-creation process. Thus, good CRM process measures provide the firm with the opportunity to gain deeper insights into how these intermediate process measures link to downstream firm performance. Several articles in this issue show these links. Thus, we do not believe that every article must focus its attention on the most obvious downstream measures of performance (e.g., profits, shareholder value). However, it is clear that more work must be done to establish the links between the many process measures that come from CRM systems and these downstream measures as well as the implied return on different CRM investments. The work of Gupta, Lehmann, and Stuart (2004) offers an excellent first step in this direction. This area should be of particular interest to researchers who want to demonstrate the link between marketing activities and shareholder value (Srivastava, Shervani, and Fahey 1998). In general, CRM creates the potential for firms to begin to treat as firm investments what were previously considered marketing costs (Rust, Lemon, and Zeithaml 2004). Furthermore, this implies that marketing could regain a central role in managing a key asset of the firm, namely, the customer asset.

Finally, even though we consider Payne and Frow's (2005) article conceptual rather than empirical, we note its relevance to the CRM-performance link. In particular, this conceptual framework emerges as the "best practice" from interaction research with several firms. If a firm does not achieve the desired results from its CRM activities, it might compare its practices with the best practice template that Payne and Frow provide. This comparison could reveal gaps in how the firm implements CRM relative to best practices.

Having said this, we note that Payne and Frow's (2005) framework is largely silent about how a particular context or process might interact with another process to produce differential results from CRM activities. Although the articles published in this special section show that CRM activities lead to enhanced firm performance, they also reveal situations in which CRM activities have more or less positive effects on firm performance. This leads to our next proposition:

P₄: Holding fixed the level of CRM investment, the effectiveness of CRM activities depends on how CRM is integrated with the firm's (a) existing processes and (b) preexisting capabilities.

We previously noted that CRM activities need to be integrated into the fabric of the overall operations of the firm. Because different firms have different core capabilities, it is not surprising that CRM activities have a differential effect depending on the context of where and when they are implemented. This is similar to what has been observed in the context of relationship management (e.g., Coviello et al. 2002). Specifically, Jayachandran and colleagues (2005) show that the positive effects of investments in CRM technology are enhanced when the firm already has the appropriate relational information processes in place. Srinivasan and Moorman (2005) show that for online retailers, the firm's strategic commitments in terms of prior bricks-and-mortar experience and online experience affect the impact of online CRM investments on the firm's performance. Notably, they also find a few cases in which increased investments in CRM are associated with negative returns in performance.

Likewise, Mithas, Krishnan, and Fornell (2005) show that CRM activity returns are enhanced when firms share information with their supply chain members. Furthermore, Thomas and Sullivan (2005) show that an enterprise CRM system that coordinates and integrates data from different channel sources enables the firm to gain new knowledge about each customer and thus enhance firm performance.

We expect that there are many other contexts in which CRM activities are either enhanced or reduced. We discuss this issue in more depth in some of the subsequent propositions. However, before doing so, we offer another proposition that may sound somewhat contradictory to the previous proposition:

P₅: Effective CRM implementation does not necessarily require sophisticated analyses, concepts, or technology.

After reading numerous submissions for this issue, we were struck by the "simplicity" of the application articles. The articles used known methodologies (e.g., latent segmentation: Lewis 2005; Thomas and Sullivan 2005), relied on known conceptual issues (e.g., adverse selection in acquiring customers: Cao and Gruca 2005), and examined small pieces of the overall set of CRM activities (e.g., use of customer lifetime value: Ryals 2005) in studying the effects of CRM on performance.

Perhaps the most striking example of this is Ryals's (2005) contribution. From a research methodology standpoint, the case study approach is technically unsophisticated. Moreover, the CRM activities implemented in these case studies are simple and straightforward. The combination of these two attributes led one reviewer of this manuscript to conclude that an important implication of this article was that even simple CRM activities yield measurable benefits for a firm.

Another surprise gleaned from the application articles is the relevance of traditional market segmentation in the context of CRM activities. Some may equate CRM with the idea that every firm offer/activity should be customized for individual consumers. However, in all of the application articles, we observed the use of basic market segmentation (Cao and Gruca 2005; Lewis 2005; Ryals 2005; Thomas and Sullivan 2005), and three of the articles identify only two segments. Admittedly, these segments were not based on standard demographics but rather on detailed analyses of prior observed behavior.

Only Ryals (2005), in one of her two case studies, shows an application in which the firm treated each customer individually, and here the firm had only ten major customers. Still, what we find most germane in considering the group of application articles published herein is that despite the simplicity in the approaches, each of these applications was able to show improvements in firm performance when the firm acted strategically in terms of using customer information to create firm value.

Jayachandran and colleagues (2005) reinforce the point about simplicity and CRM effectiveness The database they used in their analysis contained a significant number of firms that had yet to implement sophisticated CRM applications. Yet these researchers showed that as long as these firms had good relational processes in place, they were able to obtain good firm performance. Thus, although the effectiveness of CRM may vary depending on the specific context of these activities, consistent with P4, it appears that the most important element of CRM implementation is for the firm to acquire customer knowledge and then use this knowledge wisely for the dual creation of value. This leads to our next proposition.

P₆: The core of CRM is the concept of dual creation of value.

Payne and Frow's (2005) Figure 1 best demonstrates "proof" of this proposition; it shows the cocreation of value as the central element in their conceptual framework. We consider the labels "cocreation" and "dual creation" of value interchangeable, but we prefer "dual creation" given instances in which firms can create value for one customer through information drawn from other customers (e.g., Amazon) rather than direct collaboration. Jayachandran and colleagues (2005) provide a deeper description of this core idea with their delineation of five subprocesses that they refer to as relational information processes. We believe that their processes, which they label "information reciprocity," "information capture," "information integration," "information access," and "information use," directly relate to the dual creation of value.

The idea that dual creation of value is at the core of CRM is also evident in all the articles that examine the firm-customer interface. For example, Cao and Gruca (2005) provide a framework whereby the firm can better limit its target market to customers who both want to hear about the firm's particular offer and qualify for that offer. As a result, the firm does not send messages to customers who are unlikely to respond, thus minimizing the disturbance to these customers. Likewise, the firm does not send messages to customers who are unlikely to qualify for the offer, thus minimizing their disappointment. The authors note that this leads to an obvious win-win situation for the firm and its customers (i.e., the dual creation of value).

In contrast, Lewis (2005) develops a pricing scheme that creates a differential value proposition for different market segments. He notes that this raises the issue of fairness and trust because the goal is for the firm to use knowledge about customers to extract more value for the firm and therefore create less value for the customer. Similarly, Ryals (2005) shows that firms reduce their attention to customers/ customer groups after they determine that they are not able to garner enough value from these groups. Thus, for certain customers, value is taken away so that firms can increase the value they receive. In a similar manner, Thomas and Sullivan (2005) examine the dual creation of value from the firm perspective. They propose a process that enables firms to migrate customers into more profitable channels. However, this article does not address the issue of whether this process creates additional customer value.

These latter studies bring to the forefront the concept of who gets the economic rents from the value creation process. Economic theory is silent on this issue, other than to maintain that neither party can be worse off. However, if CRM is implemented in a way that leads consumers to believe that they are worse off, firms can put themselves at substantial risk. Information reciprocity can break down, and consumers may ultimately choose to opt out of relationships. This leads us to consider the "dark side" of CRM or, more generally, the potential pitfalls and unknowns that firms should consider when implementing CRM activities.

In ideal circumstances, CRM and the dual creation of value expand the pie such that both consumers and firms are better off. However, the focus in CRM applications, as some of the articles in this special section demonstrate, can be on the creation of firm value. In these cases, CRM might be considered a pie-splitting mechanism, whereby the firm can learn things about customers that enable it to take a big slice of the created value. To rephrase this in economic terms, firms may use CRM activities to attempt to extract consumer surplus. This possibility makes Wright's (1986) concept of "schemer schema" relevant in the context of CRM activities. Given customers' intuitive theories about what firms are trying to do with CRM, how will customers modify their behavior? This question leads to our first proposition about firm issues in implementing CRM activities:

P₇: The successful implementation of CRM requires that firms carefully consider issues of consumer trust and privacy.

Sometimes the firm can unobtrusively collect information about the customer at the time of the transaction. Other times, the firm must rely on the customer providing this information. It is in the firm's self-interest to collect these data; as both Mithas, Krishnan, and Fornell (2005) and Jayachandran and colleagues (2005) show, firms that acquire this knowledge are more likely to have superior firm performance. However, it may not always be in the customer's self-interest to provide these data. Lewis (2005) nicely illustrates Wright's (1986) notion of schemer schema. In this application, the firm infers customer types from observed prior behavior. However, Lewis shows that some customers anticipate what a firm will do after it observes customer behavior. This leads these customers to modify their own behavior. In other words, the consumers act strategically. If consumers act strategically, this reduces the firm's share of the value creation pie, even if the firm anticipates these reactions, though anticipation on the part of the firm will reduce this reduction in share of the value pie.

In Lewis's (2005) setting, only 5% of the customers are in the "strategic behavior" segment. However, as more consumers begin to consider that firms are using CRM to act strategically, an increasing number of customers become less trusting of firm behavior and therefore act strategically in repeated interactions with the firm. The real issue from the consumers' perspective is whether they trust that firms will use their data in a way that helps the consumer. Lewis cites some fairly prominent examples of firm behavior that may reduce customer trust in those firms (e.g., attempts by Coke and Amazon to extract consumer surplus). In today's Internet environment, there is a proliferation of spyware (and spyware blockers) that has led to a distrust of the Internet shopping environment and a desire for greater consumer privacy.

Deighton (2005) suggests that issues of consumer trust could significantly undermine CRM activities. In particular, if customers lose trust in firms and believe that their data are used by firms for purposes of exploiting them, consumers will attempt to keep their data private or to distort the data. This has led, and could continue to lead, to both individually based efforts to keep data private or collectively based efforts that lead to privacy regulations. This consumer protection is certainly reasonable and appropriate. Thus, firms should think with foresight about trust and privacy implications of their CRM activities, and researchers should continue to consider these issues (e.g., Bart et al. 2005). If the firm does not adequately consider the creation of value for both their customers and themselves, they may lose access to the data required for the dual creation of value process. Stated simply, firms should not be greedy.

P₈: The successful implementation of CRM requires that firms carefully consider issues of consumer fairness.

The precursor to some of these trust issues is fairness: Do customers trust that firms will be fair in splitting the value creation pie? However, there is an additional perspective of fairness to consider. In particular, CRM activities create the potential for differential treatment of customers who interact with a firm. Reitz (2005) provides examples in which customers did not become upset by being treated differentially, even when they were on the same airline flight. However, Reitz also notes that customers have norms of what is fair and what is unfair in terms of differential treatment of customers and that it is easy for firms to cross over the line in terms of what customers consider unfair.

Feinberg, Krishna, and Zhang (2002) demonstrate the risk of crossing this line by means of differential treatment. They report that there is more switching (lower retention) for the focal customer if another customer, who is perceived to be similar to the focal customer, receives better treatment from the same firm. They find the same result for the focal customer if the similar customer receives better treatment from a competitive firm. This is similar to the result of Boulding and colleagues (1993), who find that based on what customers observe other customers receiving either from the same firm or from competitive firms, customers develop expectations about what they should receive and downgrade their perceptions of the firm if they receive less than what they believe they should. Likewise, using a regret theory framework, Inman, Dyer, and Jia (1997) emphasize the importance of forgone alternatives in terms of the valuation of the chosen alternative and future purchase decisions. Both Srinivasan and Moorman's (2005) and Gustafsson, Johnson, and Roos's (2005) articles show results that are consistent with this prior research. They find that what customers experience, both within a particular firm and compared with other firms, drives what they believe they should receive or what they believe is fair and, thus, affects the firm's performance.

We believe that much is still unknown about the standards customers use to determine whether the firm is acting fairly and the connection between perceived fairness and trust. Thus, we hope that researchers continue to examine these standards (e.g., Mazumdar, Raj, and Sina 2005). Because these issues are intimately connected with customers' willingness to provide data and their satisfaction with the firm relationship, firms must take great care in monitoring and managing customer perceptions of trust and fairness.

P₉: Inappropriate and incomplete use of CRM metrics can put the firm at risk of developing core rigidities, thus leading to long-term failure.

According to Payne and Frow (2005), one of the key components of CRM is a good measurement process. However, most of the most popular measures of current CRM systems (e.g., acquisition, retention, cross-selling, up-selling, customer migration, customer lifetime value) are outcome measures. These outcome measures are important and necessary. However, they may not be directly linked to the value dual-creation process, and as we previously noted, this is the core concept of CRM. Thus, it is essential that the firm also develops measures that are directly connected with this value dual-creation process, enabling the firm to understand the drivers of value and thus to ensure long-term success. The flip side of this is that failing to do so can lead to doing too much of a "good" thing, producing core rigidities (Atuahene-Gima 2005; Leonard-Barton 1992) and long-term failure.

A few examples may be useful to illustrate this point. An early insight in the development of CRM was the importance of customer retention. Indeed, it can be shown that lifetime value calculations are more sensitive to improvements in customer retention than customer acquisition (Heskett et al. 1994). Fuller (2005) describes the implications of this connection between lifetime value and retention at L.L.Bean, which focused on retention at the expense of acquisition activities. The consequence of this focus was short-term gains in profitability but, unfortunately, at the expense of new customer acquisition and a reduction in the long-term value of the firm. Subsequently, L.L.Bean adjusted the balance between acquisition and retention activities.

This example brings the importance of customer portfolio management to the forefront (Johnson and Selnes 2004). Thinking about balancing the customer portfolio in terms of customer needs, implied firm actions to create customer value, and how these customers create company value leads us to ponder one of the recommendations in the bank case study that Ryals (2005) reports. On the basis of lifetime value calculations, the bank decided to stop targeting younger, less profitable loan customers. This decision is potentially based on incomplete use of metrics in two different ways. First, it could be that the younger, less profitable customers evolve over time into older, more profitable customers. The lifetime value metric needs to be complete in the sense of capturing this potential dynamism in customer value (Du 2005). Second, it could be that dropping these smaller, less profitable customers undermines the economies of scale that enable the firm to generate profits from larger customers, as was the case for a large European financial services company (Johnson and Selnes 2005). In this case, the lifetime value metric needs to be complete in terms of considering potential externalities across customer groups. This requires the firm to allocate the costs associated with providing customer value accurately.

Another example to illustrate this proposition is what took place over several years at Xerox. Xerox was an early leader in customer satisfaction and one of the first winners of the Baldridge Award. It created excellent information systems to collect and diffuse detailed customer metrics to relevant parts of the organization, which resulted in increases in customer satisfaction and firm value. However, many of these systems focused on measuring customer satisfaction with existing products and services. It is possible that, over time, these measurement processes, which had been a source of strength for Xerox, became core rigidities, ultimately leading to subpar performance for the company. If key CRM metrics do not directly assess the value dual-creation process, there is a risk that the underlying value proposition that generates outcome metrics, such as satisfaction, retention, acquisition, and lifetime value, can slowly degrade.

Thus, although CRM enables the firm to obtain a large number of measures, it is important that at least some of these measures connect to the current and future value creation process. This should enhance the firm's innovation activities and, in general, keep the firm competitive over time. It also raises an interesting research question: What is the relationship between the level of the firm's CRM activities and its level of innovation?

P₁₀: Successful implementation of CRM requires that firms incorporate knowledge about competition and competitive reaction into CRM processes.

It is well accepted that CRM is a strategic initiative. As such, it should be held accountable to the same standard as the evaluation of other strategic choices that a firm faces (Boulding and Staelin 1995). Does CRM provide sustainable advantage for the firm in the face of competitive reaction? A standard issue in the assessment of returns to strategic choices is that if generalizations exist about strategic relationships, firms can be expected to compete away the advantages implied by these relationships (Wensley 1982). This issue is especially salient given that P₃ suggests that companies have developed proven CRM practices that enhance firm performance.

We find it surprising that the CRM literature and the articles in this special section are largely silent on this issue of competitive reaction. We note that Payne and Frow's (2005) five elements of CRM do not explicitly reflect competition, and none of the empirical articles directly considers the role of competition. We find this omission in the CRM literature especially surprising given that the evolution of CRM can be traced back to the market orientation literature. In this literature, there is a discussion of whether firms should be customer focused or competitor focused, and the conclusion is that firms should be market focused; that is, firms should focus on competitors, customers, and company capabilities (Kohli and Jaworski 1990). Thus, in addition to addressing the issue of sustainability in the face of competitive reaction, we can ask a more general question: How is competitor focus integrated into CRM?

We believe that a failure to integrate competition into a firm's CRM activities potentially puts it at serious risk. The biggest risk is related to the previous proposition (i.e., the destruction of the dual creation of value due to innovation from competition). For example, when innovation enables document copies to be made through distributed desktop printing rather than centralized copying, what does this imply for the dual creation of value for Hewlett-Packard and Xerox, respectively?

In summary, we believe that research is necessary in this area on two dimensions: First, work must be done that shows how to integrate competition into the processes underlying CRM. Second, work must be done that considers the conditions in which CRM yields sustainable advantage in the face of competitive reaction. With regard to this latter issue, it is difficult to imagine how the technology underlying CRM could create such an advantage. However, the relational information processes, the customer knowledge generated from these processes, the integration of processes, and the customer loyalty resulting from the value creation processes may be difficult to imitate.

P₁₁: Effective CRM implementation requires coordination of channels, technologies, customers, and employees.

Again, an interesting aspect of this special section is what is not included. Little attention is given to the role of employees in the implementation of effective CRM activities. In discussing how Continental Airlines went from worst to first in customer satisfaction, Reitz (2005) stresses the importance of firms having people issues under control before investing in expensive CRM technologies. The articles in this issue that come closest to addressing the role of people in CRM are those of Jayachandran and colleagues (2005) and Srinivasan and Moorman (2005). Jayachandran and colleagues detail the processes that are needed to collect, use, and analyze customer data effectively, but they only hint at how to ensure that employees use these processes. Srinivasan and Moorman measure employee involvement in the collection and dissemination of customer data and show that this involvement is positively associated with good firm performance. However, they also do not provide insights into how to ensure that people in firms exhibit this behavior. Thomas and Sullivan (2005) demonstrate the potential value of enterprise systems that integrate customer information across channels. However, they do not indicate how to bundle these enterprise systems with people processes. For example, it would be nice for the catalog customer service representatives to have access to customers' shopping activities across all company channels.

Because employees are an integral part of the delivery of CRM activities, we believe that the organizational issues relevant to CRM are a critical area that deserves a firm's attention. Data and technology processes and systems are critical for CRM activities, but without appropriate human interaction with these processes and systems, the returns on investments in these areas are at risk. Vorhies and Morgan (2005) make a similar point in noting the interdependence of marketing capabilities. Because little is known about how people issues connect to the success of CRM activities, we believe that this is an area worthy of researcher attention. A good starting point might be the services marketing literature. This literature recognizes that services lie at the intersection of marketing, human resources, and operations. We refer the reader to the work of Zeithaml, Bitner, and Wilson (2000) for an extensive discussion of this literature.

Finally, related to the previous proposition, we also note that the integration of CRM across both people and processes may be difficult to imitate and thus provide a source of sustainable competitive advantage. Thus, we hope that others will provide deeper insights into the conditions needed to integrate people into CRM activities successfully.

On the basis of our discussion of the previous propositions, it is clear that there are additional substantive unknowns pertaining to the implementation of CRM and the understanding of the effects of CRM activities. We now turn to more general methodological considerations. We provide six recommendations that we believe are important for scholars interested in pursuing research in the field of CRM.

R₁: CRM research should focus on the interaction among sub-processes or the interaction among processes, not total CRM systems.

We previously noted that CRM entails the integration of numerous processes; takes place across multiple areas of the value chain; and involves the confluence of technologies, data, and people. We believe that CRM is too complex and integrative to expect any one study to model all these aspects of CRM empirically. Thus, we expect that researchers will follow the lead of the eight empirical articles in this special issue and focus on specific areas within CRM, searching for insights pertaining to how particular processes or subprocesses behave and/or interact. However, we note an inherent tension in doing this. Researchers must remember that other parts of the CRM process may modify their relationships of interest. This will require researchers to either model these effects or control for them. In the following recommendations, we focus on specific methodological issues that investigators should take into account when conducting CRM research.

R₂: CRM research should have the appropriate measures available for the desired insights.

Most studies on CRM are involved in one way or another with firm performance. We previously noted that the articles in this special issue use several different firm performance measures. Ryals (2005) and Cao and Gruca (2005) use profit. Lewis (2005) uses revenues. Gustafsson, Johnson, and Roos (2005) use retention, and Srinivasan and Moorman (2005) and Mithas, Krishnan, and Fornell (2005) use satisfaction. Jayachandran and colleagues (2005) use a combination of retention and satisfaction. Ultimately, CRM is about linking firm actions to the many stakeholders' values. One of the major stakeholders of relevance to CRM is the firm owners (i.e., the shareholders). A key objective for these stakeholders is profit maximization. Thus, we hope that researchers obtain cost data whenever possible, thereby enabling them to assess profits. This may require them to address the allocation of costs directly across different business units or customers groups, as Ryals (2005) does at both the customer and the segment levels. Beyond this, we hope that further work directly examines the link between CRM activities and shareholder value.

A second key stakeholder of relevance to CRM is the customer. A key objective for customers is utility/value maximization. Here, there is a great deal of research that examines how CRM connects to this objective, using the proxy of customer satisfaction. However, we reiterate our previously stated belief that more work could be done that examines how CRM affects measures of the value creation process rather than value creation outcomes. In other words, how does CRM connect to innovation and constant renewal of value creation for customers?

Other stakeholders of relevance to CRM include employees, suppliers, and collaborators. Given the integrative nature of CRM, we hope that further research explores how measures of relevance to these different stakeholders are related to CRM activities. We expect that these measures are present in a variety of other literature streams (e.g., strategy, supply chain management, organizational behavior and design).

In addition, the CRM environment provides a rich setting to test the relationship between and among various measures. Thus, Gustafsson, Johnson, and Roos (2005) examine the uniqueness of constructs that tap satisfaction, affective commitment, and calculative commitment as they are related to retention. This work emphasizes the issue that though constructs relevant to CRM may be conceptually distinct, their effects may be empirically indistinguishable, as is found with the satisfaction and affective commitment constructs. This finding suggests that careful attention to measurement issues is required when testing subtle theoretical effects relevant to the CRM domain.

R₃: Research should provide conclusive evidence with respect to the causal effects of CRM activities.

Most CRM researchers are interested in assessing causality. One research strategy commonly found in the study of CRM (among other fields) is to obtain responses for several firms and then determine whether there is a tendency for firms of a particular type to display different behavior than firms of another type. Often, the researcher obtains these data by asking respondents to fill out a survey. It has long been known that if both the dependent variables and the independent variables come from the same respondent, it is likely that at least part of any association between these variables is due to common method bias. A way to avoid this bias is to use two different sources, one for the independent variables and one for the dependent variables. Two of the articles using cross-sectional data in this issue (i.e., Mithas, Krishnan, and Fornell 2005; Srinivasan and Moorman 2005) use this approach. The other article in this issue that uses cross-sectional data (Jayachandran et al. 2005) relies on survey data collected from only one source within a firm. However, Jayachandran and colleagues' interests center on the effects of an interaction term, something that could not be explained by common method bias. Moreover, they present theory that predicts the form of this interaction. Thus, we can be more confident in their interpretation of the findings.

Causality is difficult to determine if the only data available are cross-sectional. The obtained results are associational and subject to interpretations of reverse causality. Thus, it is highly desirable to have temporal separation between the independent and the dependent variables, as is shown by Mithas, Krishnan, and Fornell (2005), Srinivasan and Moorman (2005), and Gustafsson, Johnson, and Roos (2005).

In general, we note the relevance of longitudinal data or cross-sectional/longitudinal data for CRM research. This is especially true given policy changes that CRM research suggests. For example, Ryals (2005) shows an increase in business unit profits due to CRM, but she cannot document the validity of lifetime value estimates based on data from a single point in time. Therefore, Ryals must make assumptions about long-term profitability implications. Likewise, Thomas and Sullivan (2005) do not document before and after behavior, given a change in communication strategy that leads to channel migration. Instead, they must assume that model estimates made from a certain point in time continue to hold; that is, when customers migrate to a new channel, they assume the shopping behaviors of those already in that channel. We note that Lewis (2005) also does not conduct a longitudinal study. However, he is better able to assert what would happen if the firm takes a proposed action because his model reflects the interaction of the different players and is based on an underlying utility formulation. This enables him to make forward-looking policy statements that imply causality.

We believe that if the field is to advance, more emphasis must be placed on documenting causality. Further research should also reflect that most firm-customer interactions are dynamic and time varying. As such, CRM systems should help provide data for this research, because a key element of these systems is the collection over time of data that document critical aspects of the process. Perhaps the major issue for academic researchers who are interested in assessing causal relationships within CRM involves accessing these data along with the cooperation of firms.

R₄: Research should acknowledge that firms do not choose CRM activities in the abstract; instead, they choose these activities on the basis of market response to these activities along with other factors, such as particular firm skills and capabilities.

Customer relationship management activities are not static. Firms and customers choose (and change) their behavior on the basis of others' actions to maximize value. Ryals (2005) shows how managers modified their behavior toward individual customers after being given access to measures of customer lifetime value. Lewis (2005) documents how customers act strategically to alter their behavior depending on the actions of the firm and that firms can alter their behavior on the basis of consumer behavior. Thomas and Sullivan (2005) show how firms can modify customers' channel choice through different communication strategies. Payne and Frow (2005) emphasize the dual creation of value between customer and firm. In all these cases, the actions of one player are chosen because of or depend on the actions of other players. In addition, Jayachandran and colleagues' (2005), Srinivasan and Moorman's (2005), and Mithas, Krishnan, and Fornell's (2005) research implies that firms should perhaps choose their CRM decisions on the basis of existing resources and capabilities.

All of this provides strong evidence that CRM activities should be considered choice variables and thus treated as endogenous in empirical models. As is well known from the strategy literature, if this is not taken into account, the result may be biased estimates of the effects of CRM activities on performance. This is because unobserved determinants of these CRM choices may also be correlated with the dependent measure of interest in the research. Thus, we hope that further CRM research uses structural modeling and estimation approaches that address this issue.

We note that this concern about endogeneity is also connected to our recommendation for increased use of longitudinal or cross-sectional/longitudinal data for CRM research. Correcting for the possibility of endogeneity bias is difficult when there is only access to cross-sectional data (Boulding and Staelin 1995). Given this limitation, we find that Mithas, Krishnan, and Fornell's (2005) approach is quite creative. Because they rely on a cross-sectional database, they cannot control for the issue of endogeneity bias. However, they present sensitivity analysis that indicates the degree to which the estimate of their theoretical effect of interest changes as a function of the (unknown) correlation of unobserved factors with the independent variable of interest. Thus, although their reported estimate is not necessarily free from bias, the reader can assess the stability of this result in the face of potential bias from unobserved factors.

R₅: CRM research should suitably address potential heterogeneity in customer behavior.

An underlying premise of CRM is that customers have different needs, and thus the firm should treat them differently. By definition, this implies that researchers need to acknowledge heterogeneity in customer behavior. Given the extensive focus on heterogeneity issues in the field of marketing, there are several approaches that researchers can take to address this issue. Lewis (2005) and Thomas and Sullivan (2005) do this by constructing latent class segmentation schemes (Kamakura and Russell 1989) and then estimating different response coefficients for each segment. Cao and Gruca (2005) take a different approach by controlling for heterogeneity through observable differences in their customers. Gustafsson, Johnson, and Roos (2005) also control for heterogeneity through observable customer differences by constructing an individual-level variable that, in effect, segments customers in terms of current churn behavior based on past churn behavior. This approach is reminiscent of what Guadagni and Little (1983) did in the context of brand choice. Another approach, which none of the articles in this special issue uses, is to use Bayesian estimation (Allenby and Rossi 1999), which, at least conceptually, enables every person to have his or her own response coefficients. However, the bottom line is that CRM rests on the notion that customers may differ. Researchers need to acknowledge this potential diversity whenever they analyze customer response to firm actions.

R₆: The research results should generalize rather than be idiosyncratic to the chosen research domain.

A major feature of high-impact academic research is that the findings, concepts, and ideas that result from the research can be applied to a broad set of environments and/ or domains. There are two basic approaches that a researcher can take to achieve such a goal for CRM research. The first approach is to base the research analysis on a wide range of firms (i.e., use cross-sectional analysis). Although this enables the researcher to address the issue of generalizability directly, it opens up the issue of being able to control for many environmental factors that could affect the dependent variable of interest along with the independent variables that vary across firms. One standard approach that researchers who conduct cross-sectional analyses use is to include variables that control for these differing environmental factors in their analyses. This is the approach that Srinivasan and Moorman (2005), Mithas, Krishnan, and Fornell (2005), and Jayachandran and colleagues (2005) use.

A key issue is not whether a cross-sectional analysis has adequate control variables but whether the researcher can claim that he or she has included all relevant environmental variables. Obtaining good measures for all potentially relevant environmental measures is often difficult if not impossible. Unfortunately, if one or more of the omitted environmental variables correlate with both the independent variable of theoretical interest and the dependent measure, the obtained result is biased. A way to circumvent this issue is for the researcher to obtain multiple observations on each firm (i.e., obtain a longitudinal, cross-sectional database). This enables the researcher to hold fixed the environmental factors that are unique to each firm by conducting both within-and between-firm analyses (see Boulding and Staelin 1995).

A second approach is to conduct a series of within-firm analyses over time and then look for commonalities among these studies. Here, the emphasis is on determining changes over time. Two examples of within-firm analyses over time in this special issue are those of Ryals (2005), who studies how managers modify their behavior over time in response to relevant CRM metrics, and Gustafsson, Johnson, and Roos (2005), who study how consumers change their behavior over time as a function of their prior behavior, satisfaction, and calculative commitment to the firm. When several of these single-firm analyses have been conducted, the researcher(s) can then perform a meta-analysis and look for generalized principles. In a way, this is what we tried to do by examining the ten articles in this special issue and then coming up with our propositions.

The field of CRM has matured over the past decade. We share Rogers's (2005) sense of excitement over this CRM maturation process. Much progress has been made, as witnessed by the number of academic centers, conferences, research papers, courses, and industry attention devoted to this topic. However, as does Rogers, we note that many unanswered questions remain to be addressed.

One question frequently asked is, What is next after CRM? Given our assertion that CRM is the outcome of the continuing evolution and integration of marketing ideas and newly available data, technologies, and organizational forms, we do not forecast a discontinuous leap after CRM. Rather, we predict a continued evolution in CRM as new ideas, technologies, and so forth, are integrated into CRM activities. We hope that this article and the other articles in this special section provide the catalyst for other researchers to continue to move the theory and practice of CRM forward in this evolutionary process.

The authors thank Ruth Bolton for her insights, guidance, and support in shaping this article and the special section and the authors of the articles in the special section for their contributions. They also thank the Teradata Center for Customer Relationship Management at Duke University, the American Marketing Association, and Journal of Marketing for their support for this special section.

( n1) This conference was cosponsored by the AMA Relationship Marketing Special Interest Group.

( n2) Space constraints force us to take a 40,000-foot perspective on the field. Thus, we do not provide an exhaustive review of the CRM literature, much less the relevant marketing literature.

Allenby, Greg M. and Peter E. Rossi (1999), "Marketing Models of Consumer Heterogeneity," Journal of Econometrics, 89 (November), 57-78.

Arndt, Johan (1979), "Toward a Concept of Domesticated Markets," Journal of Marketing, 43 (Fall), 69-75.

Atuahene-Gima, Kwaku (2005), "Resolving the Capability-Rigidity Paradox in New Product Innovation," Journal of Marketing, 69 (October), 61-83.

Bagozzi, Richard P. (1974), "Marketing as an Organized Behavioral System of Exchange," Journal of Marketing, 38 (4), 77-81.

Bart, Yakov, Venkatesh Shankar, Fareena Sultan, and Glen L. Urban (2005), "Are the Drivers and Role of Online Trust the Same for All Web Sites and Consumers? A Large-Scale Exploratory Empirical Study," Journal of Marketing, 69 (October), 133-52.

Berry, Leonard L. (1983), "Relationship Marketing," in Emerging Perspectives on Services Marketing, Leonard L. Berry, G. Lynn Shostack, and Gregory Upah, eds. Chicago: American Marketing Association, 25-28.

Boulding, William, Ajay Kalra, Richard Staelin, and Valarie A. Zeithaml (1993), "A Dynamic Process Model of Service Quality: From Expectations to Behavioral Intentions," Journal of Marketing Research, 30 (February), 7-27.

----- and Richard Staelin (1995), "Identifying Generalizable Effects of Strategic Actions on Firm Performance: The Case of Demand-Side Returns to R&D Spending," Marketing Science, 14 (Summer), G222-G237.

Cao, Yong and Thomas S. Gruca (2005), "Reducing Adverse Selection Through Customer Relationship Management," Journal of Marketing, 69 (October), 219-29.

Coviello, Nicole E., Roderick J. Brodie, Peter J. Danaher, and Wesley J. Johnston (2002), "How Firms Relate to Their Markets: An Empirical Examination of Contemporary Marketing Practices," Journal of Marketing, 66 (July), 33-47.

Day, George S. (1994), "The Capabilities of Market-Driven Organizations," Journal of Marketing, 58 (October), 37-52.

Deighton, John (2005), "Privacy and Customer Management," Customer Management, (MSI conference summary). Cambridge, MA: Marketing Science Institute, 17-19.

Du, Rex (2005), "Projecting Consumers' Needs Over the Lifecycle: A Structural Approach," working paper, Fuqua School of Business, Duke University.

Dwyer, F. Robert, Paul H. Schurr, and Sejo Oh (1987), "Developing Buyer-Seller Relationships," Journal of Marketing, 51 (April), 11-28.

Feinberg, Fred. M., Aradhna Krishna, and Z. John Zhang (2002), "Do We Care What Others Get? A Behaviorist Approach to Targeted Promotions," Journal of Marketing Research, 39 (August), 277-92.

Fuller, Steve (2005), "Into the Woods and Out Again: Revitalization Through Customer Metrics," Customer Management, (MSI conference summary). Cambridge, MA: Marketing Science Institute, 15-17.

Gaski, John F. (1984), "The Theory of Power and Conflict in Channels of Distribution," Journal of Marketing, 48 (Summer), 9-30.

Grönroos, Christian (1994), "Qua Vadis, Marketing? Toward a Relationship Marketing Paradigm," Journal of Marketing Management, 10 (5), 347-60.

Guadagni, Peter M. and John D.C. Little (1983), "A Logit Model of Brand Choice Calibrated on Scanner Data," Marketing Science, 2 (Summer), 203-239.

Gummesson, Evert (1987), "The New Marketing: Developing Long-Term Interactive Relationships," Long Range Planning, 20 (4), 10-20.

Gupta, Sunil, Donald R. Lehmann, and Jennifer Ames Stuart (2004), "Valuing Customers," Journal of Marketing Research, 41 (February), 7-19.

Gustafsson, Anders, Michael D. Johnson, and Inger Roos (2005), "The Effects of Customer Satisfaction, Relationship Commitment Dimensions, and Triggers on Customer Retention," Journal of Marketing, 69 (October), 210-18.

Heskett, James L., Thomas O. Jones, Gary W. Loveman, Earl W. Sasser Jr., and Leonard A. Schlesinger (1994), "Putting the Service-Profit Chain to Work," Harvard Business Review, 82 (March-April), 164-74.

Inman, J. Jeffrey, James S. Dyer, and Jianmin Jia (1997), "A Generalized Utility Model of Disappointment and Regret Effects on Post-Choice Valuation," Marketing Science, 16 (2), 97-112.

Jayachandran, Satish, Subhash Sharma, Peter Kaufman, and Pushkala Raman (2005), "The Role of Relational Information Processes and Technology Use in Customer Relationship Management," Journal of Marketing, 69 (October), 177-92.

Johnson, Michael D. and Fred Selnes (2004), "Customer Portfolio Management: Toward a Dynamic Theory of Exchange Relationships," Journal of Marketing, 68 (April), 1-17.

----- and ----- (2005), "Diversifying Your Customer Portfolio," Sloan Management Review, forthcoming.

Kamakura, Wagner A. and Gary J. Russell (1989), "A Probabilistic Choice Model for Market Segmentation and Elasticity Structure," Journal of Marketing Research, 26 (November), 379-91.

Keller, Kevin L. (1993), "Conceptualizing, Measuring, and Managing Customer-Based Brand Equity," Journal of Marketing, 57 (January), 1-22.

Kohli, Ajay K. and Bernard J. Jaworski (1990), "Market Orientation: The Construct, Research Propositions, and Managerial Implications," Journal of Marketing, 5 (April), 1-18.

Lehmann, Donald R. (2004), "Metrics for Making Marketing Matter," Journal of Marketing, 68 (October), 73-75.

Leonard-Barton, Dorothy (1992), "Core Capabilities and Core Rigidities: A Paradox in Managing New Product Development," Strategic Management Journal, 13 (Summer), 111-26.

Levitt, Theodore (1960), "Marketing Myopia," Harvard Business Review, 38 (July-August), 45-60.

----- (1969), The Marketing Mode: Pathways to Corporate Growth. New York: McGraw-Hill.

Lewis, Michael (2005), "Incorporating Strategic Consumer Behavior into Customer Valuation," Journal of Marketing, 69 (October), 230-38.

MacNeil, Ian R. (1978), "Contracts: Adjustment of Long-Term Economic Relations Under Classical, Neoclassical and Relational Contract Law," Northwestern University Law Review, 72, 854-902.

Mazumdar, Tridib, S.P. Raj, and Indrajit Sinha (2005), "Reference Price Research: Review and Propositions" Journal of Marketing, 69 (October), 84-102.

Mithas, Sunil, M.S. Krishnan, and Claes Fornell (2005), "Why Do Customer Relationship Management Applications Affect Customer Satisfaction?" Journal of Marketing, 69 (October), 201-209.

Morgan, Robert M. and Shelby D. Hunt (1994), "The Commitment-Trust Theory of Relationship Marketing," Journal of Marketing, 58 (July), 20-38.

Narver, John C. and Stanley F. Slater (1990), "The Effect of a Market Orientation on Business Profitability," Journal of Marketing, 20 (October), 20-35.

Parvatiyar, Atul and Jagdish Sheth (2001), "Customer Relationship Management: Emerging Practice, Process and Discipline," Journal of Economic and Social Research, 3 (2), 1-34.

Payne, Adrian and Pennie Frow (2005), "A Strategic Framework for Customer Relationship Management," Journal of Marketing, 69 (October), 167-76.

Peppers, Don and Martha Rogers (1993), The One-to-One Future: Building Relationships one Customer at a Time. New York: Currency Doubleday.

Pine, Joseph B. (1993), Mass Customization. The New Frontier of Business Competition. Boston: Harvard University Press.

Reitz, Bonnie (2005), "Worst to First to Favorite: The Inside Story of Continental Airline's Business Turnaround," Customer Management, (MSI conference summary). Cambridge, MA: Marketing Science Institute, 4-5.

Rigby, Darrell K., Frederick F. Reichheld, and Phil Schefter (2002), "Avoid the Four Perils of CRM," Harvard Business Review, 80 (February), 101-109.

Rogers, Martha (2005), "Customer Strategy: Observations from the Trenches," Journal of Marketing, 69 (October), 262-63.

Rust, Roland T., Tim Ambler, Gregory S. Carpenter, V. Kumar, and Rajendra K. Srivastava (2004), "Measuring Marketing Productivity: Current Knowledge and Future Directions," Journal of Marketing, 68 (October), 76-89.

-----, Kay Lemon, and Valarie Zeithaml (2004), "Return on Marketing: Using Customer Equity to Focus Marketing Strategy," Journal of Marketing, 68 (January), 109-127.

Ryals, Lynette (2005), "Making Customer Relationship Management Work: The Measurement and Profitable Management of Customer Relationships," Journal of Marketing, 69 (October), 252-61.

Sheth, Jagdish (2005), "The Benefits and Challenges of Shifting Strategies," Customer Management, (MSI conference summary). Cambridge, MA: Marketing Science Institute, 4-5.

Srinivasan, Raji and Christine Moorman (2005), "Strategic Firm Commitments and Rewards for Customer Relationship Management in Online Retailing," Journal of Marketing, 69 (October), 193-200.

Srivastava, Rajendra K., Tasadduq Shervani, and Liam Fahey (1998), "Market-Based Assets and Shareholder Value: A Framework for Analysis," Journal of Marketing, 62 (January), 2-18.

Thomas, Jacquelyn S. and Ursula Y. Sullivan (2005), "Managing Marketing Communications with Multichannel Customers," Journal of Marketing, 69 (October), 239-51.

Vargo, Stephen L. and Robert R. Lusch (2004), "Evolving to a New Dominant Logic for Marketing," Journal of Marketing, 68 (January), 1-17.

Vorhies, Douglas W. and Neil A. Morgan (2005), "Benchmarking Marketing Capabilities for Sustainable Competitive Advantage," Journal of Marketing, 69 (January), 80-94.

Webster, Frederick E. (1992), "The Changing Role of Marketing in the Corporation," Journal of Marketing, 56 (October), 1-17.

Wensley, Robin (1982), "PIMS and BCG: New Horizons or False Dawn," Strategic Management Journal, 3 (April-June), 147-59.

Whiting, Rick (2001), "CRM's Realities Don't Match Hype," Information Week, (March 19), 79-80.

Wright, Peter (1986), "Schemer Schema: Consumers' Intuitive Theories About Marketers' Influence Tactics," in Advances in Consumer Research, Vol. 13, Richard Lutz, ed. Provo, UT: Association for Consumer Research, 1-3.

Zeithaml, V.A., M.J. Bitner, and A. Wilson (2000), Services Marketing: Integrating Customer Focus Across the Firm. Boston: Richard D. Irwin/McGraw-Hill.

~~~~~~~~

By William Boulding; Richard Staelin; Michael Ehret and Wesley J. Johnston

William Boulding is a professor and Associate Dean, Fuqua School of Business, Duke University

Richard Staelin is Edward and Rose Donnell Professor of Business Administration, Fuqua School of Business, Duke University

Michael Ehret is Assistant Professor of Marketing, Freie Universität Berlin

Wesley J. Johnston is CBIM Roundtable Professor of Marketing, Department of Marketing, Georgia State University

A Longitudinal Study of Complaining Customers' Evaluations of Multiple Service Failures and Recovery Efforts 

The authors report a repeated measures field study that captures complaining customers' perceptions of their over-all satisfaction with the firm, likelihood of word-of-mouth recommendations, and repurchase intent during a 20-month span that includes two service failures and recovery attempts. The findings suggest that though satisfactory recoveries can produce a "recovery paradox" after one failure, they do not trigger such paradoxical increases after two failures. Furthermore, "double deviations" can occur following two consecutive unsatisfactory recoveries or following an unsatisfactory recovery in response to a second failure. The findings indicate that customers reporting an unsatisfactory recovery followed by a satisfactory recovery reported significantly higher ratings at the second postrecovery period than did customers reporting the opposite recovery sequence. The outcome of the second recovery also demonstrated a significant influence on customer ratings (positively if the recovery was satisfactory, negatively if the recovery was unsatisfactory), regardless of whether the customer found the first recovery satisfactory or unsatisfactory. In addition, although the increased change in recovery expectations and failure severity ratings from the first failure to the second is more dramatic for customers who previously reported a satisfactory recovery, the increase in attributions of blame toward the firm is more pronounced for customers who previously reported an unsatisfactory recovery. Last, the results show that recovery efforts are attenuated when two similar failures occur and when two failures happen in close time proximity.

Firms can affect customer evaluations when they attempt to recover from service failures (Smith, Bolton, and Wagner 1999; Tax, Brown, and Chandrashekaran 1998). Prior research suggests that highly effective recovery efforts can produce a "service recovery paradox" in which secondary satisfaction (i.e., satisfaction after a failure and recovery effort) is higher than prefailure levels (McCollough, Berry, and Yadav 2000; Smith and Bolton 1998). However, evidence for the paradox is sparse and mixed. Smith and Bolton (1998), employing a scenario-based experiment, report that cumulative satisfaction and patronage intentions increase above prefailure levels when respondents are very satisfied with the recovery efforts. Other studies offer contrary evidence, finding that post-recovery satisfaction levels are not restored despite effective recoveries (Bolton and Drew 1991; McCollough, Berry, and Yadav 2000). Poor service recoveries have been shown to exacerbate already low customer evaluations following a failure, producing a "double deviation" effect (Bitner, Booms, and Tetreault 1990; Hart, Heskett, and Sasser 1990). Employing a qualitative critical incident technique, Bitner, Booms, and Tetreault (1990) asked respondents to recall a dissatisfactory service experience and then explain what made them feel dissatisfied. The results indicate that poor recovery efforts intensify customer dissatisfaction. Although these studies have proved informative, they have focused only on a single failure and recovery effort. Because many service relationships are ongoing, however, customers will likely experience multiple failures over the course of a relationship. Yet it remains unclear how complainants would respond to multiple failures and recovery efforts, which suggests a need for longitudinal studies examining the dynamics of complainant perceptions over time. Such studies would help scholars and managers better understand the updating processes that complainants use in evaluating service firms. Although some longitudinal studies have examined customer satisfaction and intention (e.g., Bolton and Lemon 1999; LaBarbera and Mazursky 1983; Mittal, Kumar, and Tsiros 1999; Oliver 1980), none has explored within-subject perception changes following multiple failures and recoveries. We present a 20-month longitudinal field study that investigates within-subject evaluations of overall satisfaction with the firm, word-of-mouth (WOM) recommendations, and repurchase intent at key intervals following two customer-initiated complaints and the ensuing recovery efforts. We explore between-subject mean variations over time, depending on whether customers report satisfactory or unsatisfactory recoveries. We also consider the roles of failure severity, attributions of blame toward the firm, recovery expectations, failure similarity and type, and the time between failures in the updating process.

The Influence of Multiple Failures on Complainant Perceptions

Three extant theories suggest that multiple service failures diminish paradoxical increases in customer perceptions of firms following recovery efforts and magnify double deviation dips. Prospect theory suggests that losses are weighed more heavily than gains (Kahneman and Tversky 1979; Oliver 1997), and similarly, asymmetric disconfirmation proposes that negative performances have greater influence on satisfaction and purchase intentions than positive performances do (Mittal, Ross, and Baldasare 1998). As such, several positive experiences may be needed to overcome one negative event, and customers reporting two failures may rate the firm lower despite effective recovery efforts. Likewise, Mittal, Ross, and Baldasare (1998) find that each additional unit of positive performance has diminishing value. When a second failure occurs, complainants may focus more on the negative consequences associated with the failure, because these negative perceptions are more memorable. Thus, complainants may become desensitized to satisfactory recovery efforts, thereby mitigating their positive effects. Accordingly, satisfactory recoveries may yield paradoxical gains only in the short run. (Satisfactory recoveries in this study refer to complainant ratings above the midpoint on a three-item summated scale, and unsatisfactory recoveries refer to complainant ratings at or below the midpoint on the same three-item summated scale.)

Attribution theory also suggests diminishing complainant ratings following multiple failures. Take, for example, a situation in which bank customers complain about overcharges on their statements. Given that the bank successfully resolves the complaint, attribution theory suggests that complainants may believe that the failure was unique or due to a circumstance beyond the bank's control (i.e., an unstable attribution) (Folkes 1988). In such cases, customers may feel more positive about the firm than before the failure, triggering a recovery paradox. If another failure occurs, though, complainants may discount the circumstantial attribution and instead believe that the firm consistently makes mistakes (i.e., a stable attribution). That is, when multiple failures occur, complainants will likely reevaluate their attributions. Given that, as Weiner (2000, p. 384) has argued, "one cannot logically make unstable attributions for repeated events," customers will likely infer that multiple failures are due to problems inherent to the firm. In such cases, complainants feel heightened discontent when firms do not recover satisfactorily from two failures, generating a double deviation effect. Similarly, even consistently satisfactory recoveries may have a tempered impact following multiple failures. As such, we offer three hypotheses:[ 1]

H1: Despite perceiving two satisfactory recoveries, customers reporting two failures will rate their postrecovery overall satisfaction with the firm, repurchase intent, and favorable WOM lower than their prefailure ratings for those same variables (no paradoxical increase in perceptions of the firm).

H2: Customers perceiving two unsatisfactory recoveries following two reported failures will rate their postrecovery overall satisfaction with the firm, repurchase intent, and favorable WOM lower than their ratings after the second failure for the same variables (a double deviation effect).

H3: For customers perceiving two unsatisfactory recoveries, the magnitude of the decrease in ratings in overall satisfaction with the firm, repurchase intent, and favorable WOM from postfailure to postrecovery falls more sharply after the second failure than after the first failure (heightened discontent).

The Effects of Mixed Recoveries over Time

Although we expect diminishing service recovery returns following two satisfactory recoveries, complainants still may show a preference for when a satisfactory recovery occurs in a sequence of recoveries. Research on decision making suggests that people prefer an improving series of outcomes. For example, Ross and Simonson (1991) presented subjects with hypothetical scenarios ending with a loss (e.g., win $85 and then lose $15) versus a gain (lose $15 and then win $85). The subjects strongly preferred the scenario ending in a gain. Loewenstein and Prelec (1993) similarly argue that when a timing trade-off is involved with sequential outcomes, people become more farsighted and will prefer the sequence ending in positive rather than negative outcomes. Such effects may be explained by a recency effect (Ross and Simonson 1991) and an element of prospect theory, namely, loss aversion (Kahneman and Tversky 1979). The recency effect suggests that events occurring most recently are also most salient and are weighed more heavily when people judge the overall sequence of outcomes. Loss aversion likewise suggests that when a sequence goes from a gain to a loss, people will weigh the loss more heavily, making this sequence less attractive than a sequence going from a loss to a gain.

H4: Overall satisfaction with the firm, repurchase intent, and favorable WOM ratings after the second recovery are higher for customers perceiving an unsatisfactory/satisfactory (US) sequence than for those perceiving a satisfactory/unsatisfactory (SU) sequence (ratings for US ratings for SU at post-Recovery 2).

Similar logic suggests that an unsatisfactory recovery followed by a satisfactory recovery will result in improved customer ratings over time, despite the customer reporting a prior unsatisfactory recovery. Consistent with our previous hypotheses, low ratings are likely to occur when a first failure is followed by an unsuccessful recovery. When a second failure occurs followed by a successful recovery, however, complainants will likely focus on their most recent experience and adjust their ratings upward. Thus, ratings are likely to improve when complainants experience a US recovery sequence, which is consistent with a recency effect (Ross and Simonson 1991).

We also posit that a satisfactory recovery followed by an unsatisfactory recovery will generate a double deviation effect. When complainants perceive a satisfactory recovery, they often give the firm higher ratings. However, these complainants are also likely to update their expectations upward (Grayson and Ambler 1999). Prospect theory and asymmetric disconfirmation theory suggest that negative performances influence customer affect more than positive performances. Complainants experiencing two negative events (second failure and unsatisfactory recovery) following a satisfactory first recovery likely weigh the negative events more heavily than the satisfactory recovery, which results in significant rating dips. Such dips are also consistent with a recency effect (Ross and Simonson 1991).

H5: Despite perceiving an unsatisfactory (satisfactory) first recovery, customers perceiving a satisfactory (unsatisfactory) second recovery, that is, a US sequence (SU sequence), will report increases (decreases) in their ratings of overall satisfaction with the firm, repurchase intent, and favorable WOM from post-Failure 2 to post-Recovery 2 (a recency effect).

Expectations, Contextual Influences, and the Downside of Service Recovery

Service recovery expectations. Consistent with prior research, we conceptualize service recovery expectations as customers' predictions regarding the extent to which a firm will handle their complaint (Boulding et al. 1993; Oliver 1997). Some researchers assert that postcomplaint handling evaluations increase when expectations are met or exceeded (Tax and Brown 1998), whereas other researchers suggest that expectations increase over time (Boulding et al. 1993; Grayson and Ambler 1999). However, research has not explored how customers in ongoing service relationships update their recovery expectations after reporting multiple failures. Given that negative events are salient and easily recalled, customers previously reporting one failure will likely consider their prior experience when predicting what to expect after a second failure. Complainants are more likely to attribute one failure to bad luck or causes outside the firm's control and expect only moderate redress. When another failure occurs, however, they are likely to attribute a stable pattern of failures to the firm (Weiner 2000). To the extent that this pattern is blamed on the firm, complainants will expect more extensive redress after the second failure than after the first.

Furthermore, although complainants perceiving satisfactory recoveries may rate the firm higher for its efforts (e.g., recovery paradox), they are also likely to view the solid performance as a signal to adjust future recovery expectations upward. Such adjustments are theoretically consistent with forward assimilation, in which expectations become consistent with satisfaction (Oliver and Burke 1999). Conversely, expectations should not rise as much for complainants who previously perceived an unsatisfactory recovery, partly because the past experience offers a cue that future recoveries may also be weak. Thus, complainants perceiving a satisfactory recovery after a first failure will hold higher recovery expectations for a second failure. Accordingly, the downside of recovering well is managing higher expectations in the future.

H6: Customers reporting two failures have higher recovery expectations for the second failure than for the first failure.

H7: The magnitude of the increase in recovery expectations from the first to the second failure is greater for customers perceiving a satisfactory first recovery than for customers perceiving an unsatisfactory first recovery.

Failure severity. Customer evaluations decline as service failures become more severe (Smith, Bolton, and Wagner 1999). However, what happens to severity perceptions when two failures are reported? Because one unit of loss is more salient than one unit of gain, customers may more easily recall a failure incident than the recovery effort that followed. It seems evident that complainants experiencing a second failure will have inflated perceptions of severity when they still recall the first failure (Seiders and Berry 1998). Although the recovery paradox suggests that satisfactory recoveries enhance complainant evaluations, the service firm must then manage the higher expectations that are likely to follow. From an attribution perspective, customers reporting two failures may sense a pattern of negative performances. As such, customers may recall the first failure and combine its losses with those they perceive following the second failure. These complaining customers may begin viewing the failures as stable problems inherent to the firm. Furthermore, customers may perceive the sequential failures as one overall failure, ultimately heightening the severity of the second failure. Previously unsatisfied complainants will likely report higher severity perceptions because of the magnified discontent stemming from experiencing two failures. However, we expect that previously satisfied complainants will report greater increases in perceived failure severity than previously unsatisfied complainants, partly because of their higher expectations.

H8: Customers reporting two failures will rate the second failure as more severe than the first failure.

H9: The magnitude of the increase in the perceived severity of the failure from the first to the second failure is larger for customers perceiving a satisfactory first recovery than for customers perceiving an unsatisfactory first recovery.

Attributions of blame. Customers engage in causal thinking to ascertain why a failure occurred (Weiner 2000). Attributions of blame, which we define as the extent to which customers hold the seller responsible for a failure, can be instrumental in shaping responses to failures. Researchers conclude that some complainants blame firms for failures even when the firm is not actually responsible (Folkes and Kotsos 1986), and complainants who believe that firms are responsible for failures will be more likely to expect redress (e.g., discounts, apologies, refunds). For example, Folkes (1984) asked respondents to recall a recent restaurant experience when they were unsatisfied with the taste of their food or beverage and to explain why they were unsatisfied. The results showed that attributions of blame toward the restaurant strongly influenced whether customers believed that they deserved apologies and refunds.

Our research extends this literature by examining how blame attributions change when more than one failure is reported. When one failure is reported and the firm responds well, complainants may attribute the failure to a circumstantial cause or consider it a distinct occurrence (unstable attribution). If the firm has multiple failures, complainants may attribute the failures to causes that are consistent and stable to the firm. It follows that attributions of blame toward the firm will increase after multiple failures are reported. We also contend that attributions will be more pronounced among complaining customers who previously perceived an unsatisfactory recovery than for those who previously perceived a satisfactory recovery, partly because they now perceive multiple problems (i.e., two failures and one unsatisfactory recovery) that are no longer considered inconsistent and "unstable."

H10: Customers reporting two failures will attribute blame for the failures to the firm more strongly after the second failure than after the first failure.

H11: The magnitude of the increase in blame attributions from the first to the second failure is larger for customers who report an unsatisfactory first recovery than for customers who report a satisfactory first recovery.

Lag between failures. Failures occurring over a short period are likely to affect complainants' perceptions more negatively than failures separated by longer periods of time. Given the prospect theory view that losses are weighed more heavily than gains, it may take several positive experiences to temper the effects of a single failure. Customers reporting two failures without a considerable time frame filled with satisfactory experiences will likely perceive higher discontent and lower ratings at the time of the second complaint (Seiders and Berry 1998). Similarly, complainants reporting two failures within a relatively short time period will easily recall the first failure. These customers may view the two failures as one larger failure, likely creating more demanding customers and possibly mitigating satisfactory recovery efforts. Given that the median time interval between failures in our study was four months, we classify a "short" time interval as four months or less and a "longer" time interval as five months or more.

H12: Customers reporting two failures within a relatively short time interval have lower post-Failure 2 ratings for overall satisfaction with the firm, repurchase intent, and favorable WOM (compared with customers who report two failures over a longer time interval). H13: Customers reporting two failures within a relatively short time interval have lower post-Recovery 2 ratings for overall satisfaction with the firm, repurchase intent, and favorable WOM (compared with customers who report two failures over a longer time interval).

Failure similarity. Not only can multiple failures lead to consumer discontent, but also this discontent can be magnified when the same failure occurs (Seiders and Berry 1998). From an attribution theory perspective, similar failures may lead complainants to believe that the firm consistently makes the same errors without improving--a stable internal attribution toward the firm. In addition, consumers reporting two similar failures are more likely to hold the firm responsible for making consistent mistakes, making it more difficult for firms handling two similar complaints to recover well. Conversely, when two different failures occur, complainants may be more likely to either attribute the failures to circumstantial, nonfirm factors or view them as distinct anomalies and thus may not evaluate the firm as harshly.

H14: Customers reporting two similar failures have lower post-Failure 2 ratings for overall satisfaction with the firm, repurchase intent, and favorable WOM (compared with customers who report two different failures).

H15: Customers reporting two similar failures have lower post-Recovery 2 ratings for overall satisfaction with the firm, repurchase intent, and favorable WOM (compared with customers who report two different failures).

Sample, Procedures, and Measures

We conducted a repeated measures (RM) field study with bank complainants across a 20-month time span. We focused on customers who registered complaints about their banking experiences at one of 116 branches of an industry-leading bank. At four time periods, respondents completed surveys that assessed perceptions from six time periods: pre-Failure 1 (i.e., T1a); post-Failure 1 (i.e., T1b); post-Recovery 1, approximately two weeks after the first recovery effort (i.e., T2); pre-Failure 2 (i.e., T3a); post-Failure 2 (i.e., T3b); and post-Recovery 2, approximately two weeks after the second recovery effort (i.e., T4). As in other behavioral research involving imperfect correlations, the repeated measures aspect of our design may generate some degree of regression toward the mean. Figure 1 offers a time line of measurement for all constructs in the study, and the measurement procedures are described subsequently.

Time Period 1: pre-Failure 1 (T1a). Upon complaining for the first time to any of the branch offices, 1356 customers were asked to participate in the study at Time Period 1. Complainants completed the T1a survey in the bank shortly after registering the complaint. During Time Period 1, bank service agents informed customers that the purpose of the study was to improve the bank's service efforts and that the study consisted of several parts. When customers agreed to fully participate in the study, the service agent distributed the pre-Failure 1 survey (T1a), asking customers to think retrospectively about all their experiences with the bank up to the recent service failure (i.e., past perceptions excluding the service failure). These experiences may have included past banking service availability, support, services offered, ease of use, customer service, and so forth. Customers were then instructed to rate prefailure overall satisfaction with the firm, repurchase intent, and favorable WOM likelihood. These constructs were measured with three, four, and three items, respectively, drawn from the extant literature (e.g., Cronin and Taylor 1994).

Time Period 1: post-Failure 1 (T1b). After completing the prefailure part of the survey (T1a), the 1356 customers were then asked to think about all their experiences with the bank up until that moment. This post-Failure 1 part of the survey (T1b) asked customers to rate their perceptions of overall satisfaction with the firm, repurchase intent, and WOM after experiencing a service failure, using items identical to those used in the T1a survey. In addition, the T1b survey asked respondents to rate their perceptions regarding service recovery expectations, attributions of blame, and failure severity. A four-item recovery expectation measure, adapted from McCollough, Berry, and Yadav (2000), asked respondents to rate the extent to which they expected the firm to effectively recover from the failure. A four-item attribution measure asked respondents to indicate the extent to which the firm was responsible for the failure, and a three-item failure severity measure asked customers to indicate the severity of the failure they reported. Respondents also provided some demographic information.

Time Period 2: post-Recovery 1 (two weeks after recovery, T2). In Time Period 2, the same measures of overall satisfaction with the firm, repurchase intent, and WOM were gathered along with a three-item satisfaction with service recovery measure (Cronin and Taylor 1994). This T2 survey was administered to customers who completed both T1a and T1b and was mailed one week after the bank concluded its recovery efforts (with hopes of reaching the customer within two weeks after recovery). The bank offered incentives to participate, and research assistants telephoned customers as a reminder to respond. Of the surveys mailed, 692 usable responses were collected and matched to T1a and T1b--a 51% response rate for Failure/Recovery 1.

Second failure and recovery data collection. Customers who reported a second failure were asked to complete surveys representing perceptions and recovery efforts for the second failure. The mean time lapse between the two failures was 6.63 months, and approximately 75% of the customers who reported a second failure did so within nine months. Of the 692 respondents who completed all surveys involving the first failure, 312 complained to the bank about a second failure. Of those, 255 completed all portions of the study across four time periods. These 255 constituted the sample used in our study. The interviewing schedule; data collection procedures; and measures for pre-Failure 2 (T3a), post-Failure 2 (T3b), and post-Recovery 2 (T4) mirrored the three surveys involving the first failure and recovery effort.

Time Period 3: pre-Failure 2 (T3a). At Time Period 3, respondents who registered a second complaint completed the second prefailure survey inside the bank. The T3a survey asked customers to think retrospectively about all their banking experiences with the bank up until the most recent service failure (i.e., past perceptions). Respondents then rated their overall satisfaction with the firm, favorable WOM likelihood, and repurchase intent prior to the second failure. To validate our T3a pre-Failure 2 retrospective measures, we compared the raw mean scores of the post-Recovery 1 measures and the corresponding pre-Failure 2 measures. There were no significant differences (t-values ranged from -.281 to .864). As such, the post-Recovery 1 means, collected on average 6.63 months prior to our pre-Failure 2 measures, did not differ from our pre-Failure 2 measures.

Time Period 3: post-Failure 2 (T3b). Also in Time Period 2, after completing the prefailure survey, respondents were asked to think about all their experiences with the bank, including the most recent service failure. The post-Failure 2 survey (T3b survey) asked customers to rate their current perceptions of overall satisfaction with the firm, repurchase intent, and WOM. The T3b survey also asked customers to rate their perceptions regarding service recovery expectations, attributions of blame, and failure severity after Failure 2.

Time Period 4: post-Recovery 2 (two weeks after recovery, T4). In the fourth time period, the second postrecovery survey gathered measures of overall satisfaction with the firm, repurchase intent, WOM, and satisfaction with service recovery (identical to T2). The second postrecovery survey (T4 survey) was mailed to customers who completed all five previous portions of the study, which resulted in our sample size of 255. All items across all surveys were measured with seven-point scales and are shown in Appendix A. The raw means and standard deviations for all measures, as well as the correlations among measures, are shown in Appendix B. Across all surveys, coefficient alpha estimates for all measures ranged from .83 to .97.[ 2]

We also collected data regarding the type of failure. Consistent with other service research, bank representatives logged failures as either "core" failures or "process" failures (Gilly and Gelb 1982; Smith, Bolton, and Wagner 1999). Core failures refer to monetary-oriented complaints that involve a problem with the product offering (e.g., incorrect account postings, overcharges, faulty overdraft protection). Of the 223 core failures reported (i.e., 88 at Failure 1 and 135 at Failure 2), 43% involved nonsufficient funds over-draft fees, 27% involved incorrect account postings, and 16% involved interest or automated teller machine over-charges. Recovery strategies for these failures included waiving some or all of the questioned fees, accurately adjusting account balances, and offering conscientious customer service (e.g., listening, empathizing, apologizing). Process failures were defined as problems with the way the bank provided the service (e.g., procedures, personal inter-actions). Of the 287 process failures (167 at Failure 1 and 120 at Failure 2), 37% involved queuing/waiting times or processes, 32% involved policies and procedures that restricted on-site banking access (versus electronic access) to low-volume customers, and 26% involved poor customer service (e.g., discourteous employees). Recovery strategies for these complaints focused on offering customers flexible and accommodating options, making policy or procedural exceptions, and providing caring personal interactions.

The sample exhibited the following demographic characteristics: 56% of the respondents were women, 42% were between 35 and 42 years of age, and 65% held college degrees; 66% of the sample had used the bank's services for at least one year. In addition, the initial complaint (i.e., Failure 1) in this study represented the first complaint recorded by the bank for each respondent, creating a baseline for accurately tracking customers' perceptions regarding their first and second complaint experiences.

Checks for Respondent and Measure Bias

We employed three checks to assess respondent and measure bias. First, the bank provided us with a sample of 316 complainants who did not participate in the study. No significant differences were found among age, sex, type of complaint, account value, or length of relationship between our study respondents and this sample. Second, we collected data from a sample of 276 bank customers who had not reported a failure. The demographic profiles of these 276 noncomplainants were not significantly different from the profiles in our study's sample, not only across data collected in our survey (i.e., age, sex, education, and length of relationship) but also across data collected by the bank (e.g., account value, types of services composing the account portfolio, customer profitability). Furthermore, the noncomplainants' ratings did not differ significantly (p > .10) from our complainants' ratings of overall satisfaction with the firm, repurchase intent, and WOM at T1a (i.e., the pre- Failure 1 ratings). Our complainant sample's postfailure ratings (T1b) of overall satisfaction with the firm, repurchase intent, and WOM were lower than the noncomplainants' ratings on these variables (p < .05).

Third, given that our T1a measure of overall firm satisfaction was retrospective, we compared it with an actual pre-failure satisfaction measure collected by the bank. The bank periodically administered a customer satisfaction survey. The bank's database showed that 97 of our 255 study participants had completed a firm-derived satisfaction measure four months before our study and before they reported any failure. The satisfaction measure stated the following: "Please rate your overall experience with [firm name] bank." We measured responses using a five-point scale anchored by "unpleasant" and "completely satisfactory." The correlation of this measure with our prefailure overall firm satisfaction measure was .91. Furthermore, we calibrated our measure such that it had five scale points, making it similar to the bank's measure. The difference between our calibrated measure and the bank's measure was not significant for the n = 97 subsample (mean difference = .03, t = .52, p > .60). In summary, these data checks suggest that rating biases due to respondents or retrospective measures were minimal.

Classification Factors

Before testing our hypotheses, we constructed quantification factors as independent variables in our analyses (Neter et al. 1996). We used the satisfaction with service recovery measure (captured at T2 and T4) to form a two-level variable(i.e., satisfactory and unsatisfactory recovery). We created this between-subjects factor by summing the scores on the items in the scale and then splitting the scores at the scale midpoint into two groups: one perceiving an unsatisfactory first recovery and another perceiving a satisfactory first recovery. Scores for the unsatisfactory group ranged from 3 to 12 (on a 21-point summated scale) and from 13 to 21 for the satisfactory group. We also split the scale for satisfaction with recovery regarding the second recovery at the scale midpoint (i.e., 3 to 12 for the unsatisfactory group and 13 to 21 for the satisfactory group).[ 3]

Data Checks

Before testing the hypotheses, we examined whether the recovery paradox and the double deviation effect existed after one failure and recovery effort. We used RM MANCOVA (multivariate analysis of covariance) with one three-level within-subjects factor (time: pre-Failure 1, post-Failure 1, and post-Recovery 1), one between-subjects factor (recovery: unsatisfactory, n = 112, and satisfactory, n = 143), and three covariates (i.e., recovery expectations, attributions of blame, and severity of Failure 2 compared with Failure 1). The objective of this data check was to investigate whether the paradox and double deviation hold following one failure and satisfactory recovery.

After controlling for the variance attributed to the covariates, we used linearly independent planned comparisons, adjusting for experiment-wide error rate, to compare estimated marginal means. (The effects for all covariates were significant and are available on request.) Our results show that customers reporting a satisfactory recovery rated their postrecovery overall satisfaction (mean = 16.03), repurchase intent (mean = 21.97), and WOM (mean = 15.42) significantly higher than their prefailure ratings for these same variables (satisfaction mean = 13.41, repurchase intent mean = 20.14, WOM mean = 9.74; Wilks' λ = .555, F = 66.17, p < .01), with a large effect size (η[sup2] = .45). The univariate effects for these variables were also significant (p < .01), fully supporting the service recovery paradox for one failure and recovery. Planned comparisons also indicated that customers perceiving an unsatisfactory recovery did not rate postrecovery overall satisfaction (mean = 9.23) and repurchase intent (mean = 16.53) significantly below their postfailure ratings for these variables (satisfaction mean = 8.60, repurchase intent mean = 16.63). Indeed, these customers rated postrecovery WOM likelihood (mean = 8.21) significantly above their postfailure ratings (mean = 6.70) following an unsatisfactory recovery, and this increase drives multivariate significance (Wilks' λ = .955, F = 3.88, p < .01, η[sup2] = .05). As such, the double deviation effect did not occur after one failure and recovery. Note, however, that the postfailure ratings may be susceptible to order effects, because they were collected sequentially in the same questionnaire with prefailure measures. Nonetheless, these data check results offer robust estimates, as they accounted for the effects of attributions of blame, failure severity, and recovery expectations.

To examine H 1-H5, we incorporated the history of the first failure into the model. We conducted multiway RM MANCOVA with two within-subjects factors, ( 1) time: prefailure, postfailure, and postrecovery and ( 2) failure: Failure 1 and Failure 2. We also had two between-subjects factors, ( 1) Recovery 1: unsatisfactory and satisfactory and ( 2) Recovery 2: unsatisfactory and satisfactory, and six covariates (i.e., recovery expectations, attributions of blame, and failure severity involving Failures 1 and 2). (With the exception of failure severity at Failure 2, all covariate effects were significant.) As is shown in the top portion of Table 1, post-Recovery 2 means across the dependent variables significantly decreased below pre-Failure 2 levels for customers perceiving two satisfactory recoveries (Wilks' λ = .706, F = 33.77, p < .01, η[sup2] = .29), supporting the assertion in H1 that the recovery paradox does not occur following two failures.

The second portion of Table 1 shows the results for H2 and H3. Post-Recovery 2 means significantly decreased below post-Failure 2 levels for customers who perceived two unsatisfactory recoveries (Wilks' λ = .623, F = 49.10, p < .01, η[sup2] = .38). This supports the assertion in H2 that the double deviation effect occurs after two failures and unsatisfactory recoveries. To test H3, we computed contrasts in accordance with Vonesh and Chinchilli's (1997) recommendations to determine whether the marginal mean decrease from postfailure to postrecovery changes from Failure 1 to Failure 2. As Table 1 shows, this analysis supports H3. The mean decrease from postfailure to postrecovery was more pronounced after the Failure 2 and two sequential unsatis-factory recoveries (Wilks' λ = .730, F = 29.97, p < .01, η[sup2] = .27).

The results for H4 and H5 are shown in the third and fourth portions of Table 1. We tested H4 by comparing the post-Recovery 2 means estimated in the preceding model between the US recovery sequence and the SU sequence. As Table 1 shows, customers reporting the US sequence rated the bank significantly higher than did customers reporting the SU sequence, in support of H4. In addition, post-Recovery 2 means for customers reporting the US sequence significantly increased above post-Failure 2 levels across the dependent variables collectively (Wilks' λ = .785, F = 22.17, p < .01, η[sup2] = .22), in support of H5. However, this increase was not significant at the univariate level for repurchase intent (F = 3.55, p > .06). Post-Recovery 2 means for customers reporting the SU sequence significantly decreased below post-Failure 2 levels across the dependent variables collectively (Wilks' λ = .822, F = 17.59, p < .01, η[sup2] = .18), in further support of H5.

To test H6 through H11, we estimated an RM multivariate analysis of variance model with one within-subjects factor (failure: Failure 1 and Failure 2), one between-subjects factor (Recovery 1: satisfactory and unsatisfactory), one blocking factor (failure type: core and process failures), and three dependent variables (i.e., recovery expectations, attributions of blame, and failure severity). After controlling for the variance explained by failure type, we were able to clarify the mean differences due to satisfaction and failure levels. The top portion of Table 2 shows means and mean differences relevant to H6-H11, and the bottom portion offers univariate statistics. H6 posits that recovery expectations significantly increase from Failure 1 to Failure 2. The expectations mean in the top portion of Table 2 indicates that expectations were significantly higher following Failure 2 (F = 49.84, p < .01, η[sup2] = .17), in support of H 6. This increase was also significantly greater for customers who perceived a satisfactory recovery to Failure 1 (F = 3.97, p < .05), in support of H7. Similarly, the extent to which customers perceived their failure as severe significantly increased over failures (F = 30.40, p < .01, η[sup2] = .11), and this increase was larger for customers who perceived a satisfactory recovery to Failure 1 (F = 5.78, p < .02, η[sup2] = .02). Thus, H8 and H9 were supported.

Table 2 also shows that H10 and H11 were supported, indicating that attributions of blame significantly increase from one failure to the next (F = 47.47, p < .01, η[sup2] = .16), and the effect is larger for customers who perceived an unsatisfactory recovery after the first failure (F = 16.22, p < .01, η[sup2] = .06).

To test H12 and H13, we again constructed a quantitative classification variable (Neter et al. 1996). We asked customers to indicate the number of months they had patronized the bank. The two measures were subtracted (months[second complaint] - months[first complaint]) to form a difference score representing the interval between failures. We then verified these self-report measures using the bank's database. Next, we used a median split to divide the customers into two groups: one reporting two failures in ≤ four months (n = 128) and another reporting two failures in ≥ five months (n = 127). We then used RM MANCOVA with one within-subjects factor (time: post-Failure 2 and post- Recovery 2), one between-subjects factor (number of months between failures: &;le; four and ≥ five), three dependent variables (overall satisfaction, repurchase intent, and WOM), and one covariate (postrecovery satisfaction) to test H12. (The covariate was significantly correlated with the dependent variables and was not significantly correlated with the independent variable, so we deemed it appropriate for this analysis.) We calculated linearly independent planned comparisons to determine whether postfailure and postrecovery means were lower for complainants who reported two failures in ≤ four months.

The top portion of Table 3 shows that postfailure means for the group that perceived two failures relatively close together were not significantly lower at the multivariate (Wilks' λ = .973, F = 2.32, p > .08) or univariate (p-values for all three variables ≥ .10) levels, which does not support H12. After incorporating another covariate (i.e., post-Recovery 2 satisfaction) into the model to control for customers' satisfaction with the second recovery, we used the previous model to test H13. (The covariate was significantly correlated with the dependent variables and was not significantly correlated with the independent variable.) The second portion of Table 3 also shows that the postrecovery means for the group that perceived two failures close together were significantly lower at both the multivariate (Wilks' λ = .312, F = 183.97, p < .01, η[sup2] = .69) and univariate (p-values for all three variables < .01) levels. As such, H13 is supported.[ 4]

To test H14 and H15, we constructed another quantitative classification factor to use as the independent variable. We obtained data from bank officials that indicated whether the two failures were similar or different. (This approach included the classification of a core or process failure.) We then created a dummy variable, where 1 = different failures and 2 = similar failures. Next, we divided customers into two groups, one reporting two similar failures (n = 118) and one reporting two different failures (n = 137), and used RM MANCOVA with one within-subjects factor (time: post-Failure 2 and post-Recovery 2), one between-subjects factor (failure type: different and similar), three dependent variables (overall satisfaction, repurchase intent, and WOM), one covariate (post-Recovery 1 satisfaction), and one blocking factor (failure type: core or process). The covariate and blocking factors were significantly correlated with the dependent variables and were not significantly correlated with the independent variable. As the third portion of Table 3 shows, postfailure means for the group that reported two similar failures were not significantly lower at either the multivariate (Wilks' λ = .984, F = 1.31, p > .27) or the univariate (p-values for all three variables > .07) level, which does not support H14. However, as shown in the bottom portion of Table 3, postrecovery means were lower for customers who reported two similar failures (multivariate: Wilks' λ = .886, F = 10.67, p < .01, η[sup2] = .11; univariate p-values for all three variables < .01), in support of H15.

By including failure type as a blocking factor, we were able to reduce the sum of squares due to error, refine our estimates, and uncover some notable findings. Respondents reporting two similar core failures (CC sequence) had significantly higher ratings after the second recovery than did those reporting two similar process failures (PP sequence) (Wilks' λ = .794, F = 21.41, p < .01, η[sup2] = .21). In addition, respondents reporting a process failure followed by a core failure (PC sequence) had significantly higher ratings after the second recovery than did those reporting a core failure followed by a process failure (CP sequence) (Wilks' λ = .670, F = 40.49, p < .01, η[sup2] = .33).

The purpose of our study was to examine complaining customers' perceptions of two service failures and recovery efforts. We summarize our results and implications as follows:

Recovery paradox: For a single failure and satisfactory recovery, customers rated the firm paradoxically higher on satisfaction, WOM, and repurchase intent. However, customers reporting another failure did not rate the firm higher despite satisfactory recoveries. Thus, despite effective recovery efforts, paradoxical increases diminish after more than one failure. Although managers should strive to recover well from mistakes, they would be ill advised to use satisfactory recoveries as a crutch for poor service. Our results suggest that firms cannot merely become recovery experts and need to get it right the first time. Firms also need to learn from their mistakes when they do fail and get it right the second time.

Double deviation effect: Although ratings of satisfaction, WOM, and repurchase intent declined after one failure, the declines were not compounded after an unsatisfactory recovery; that is, there was no double deviation effect. It seems that customers discount the effects of one failure when the firm has typically provided satisfactory service. However, when two unsatisfactory recoveries occur, the double deviation effect is strong. Customers may tolerate one unsatisfactory recovery, but they likely will not tolerate two.

Mixed recovery sequences: Customers reporting a US sequence reported higher post-Recovery 2 ratings than did those reporting an SU sequence. Furthermore, ratings from post-Failure 2 to post-Recovery 2 increased for those reporting a US sequence (and decreased for those reporting an SU sequence). Our study uncovers a potential recency effect when customers report inconsistent recovery efforts, suggesting a "what have you done for me lately?" response. In ongoing relationships in which customers likely experience multiple failures and recoveries, firms may improve previously low ratings associated with an unsatisfactory recovery by subsequently providing satisfactory recoveries. Also, although customers may tolerate an unsatisfactory recovery when it occurs after they report their first failure, they are not likely to tolerate an unsatisfactory recovery when it occurs after a second failure, even if the previous recovery was satisfactory.

Preferences for recovery sequences: Our study unveils a hierarchy of postrecovery ratings when customers report various recovery sequences. The route to the highest postrecovery ratings after two complaints is an SS sequence, followed by a US, SU, and UU sequence, respectively. As such, the past seems important only when customers recall consistent recovery efforts. When inconsistent efforts occur, the past may be important only to the extent that it helped shape prefailure ratings.

Recovery expectations: Our results show that customers adjust their expectations higher from one failure to the next. This increase was greater for customers who previously reported a satisfactory recovery. These results suggest that perhaps "no good deed goes unpunished," highlighting a potential downside of recovering well. To the extent that satisfied customers rate the firm higher and correspondingly adjust their future expectations, they may be more likely to experience dissatisfaction if the supplier fails again. Therefore, managers must carefully govern these newly enhanced service expectations.

Failure severity: Our results show that customers reporting a second failure rated the second failure more severely than they rated the first. Perhaps severity ratings are stronger when customers perceive a second failure because customers consider "failure history" rather than the individual failure at hand. Our results also demonstrated that failure severity ratings increase more among customers who formerly reported a satisfactory recovery than among those who previously reported an unsatisfactory recovery, which potentially under-scores another downside of recovering well. Because severe failures require greater effort on the part of the firm, managers may need to offer additional redress accordingly.

Attributions of blame: Our results show that when multiple failures occur, customers are likely to attribute the failures in a stable, internal manner to the firm. Customers formerly reporting unsatisfactory recoveries blame firms more than do once-satisfied customers when a second failure arises. To the degree that these customers believe that multiple failures and poor recoveries represent a pattern that is stable to the firm, they may attribute failures internally to the firm and therefore require more extensive recovery efforts.

Lags between failures: Complainants reporting two failures within a short time period did not rate the firm lower after the second failure than did those reporting two failures separated by a longer time period. It appears that two failures, regardless of the time lag between them, produce unsatisfied customers. Perhaps customers experiencing longer gaps remain focused on the failure and compress the time lag (see Hornik 1984). However, complainants rate firms lower after the second recovery when two failures occur within a shorter time period. This may make it more difficult for firms to recover when two failures occur close together, partly because customers may not have time to forget about the first failure.

Failure similarity: Complainants reporting two similar failures did not rate the firm lower than did those reporting two distinct failures, which suggests that two failures, regardless of their similarity, make customers equally unsatisfied. However, failure similarity affects customer responses to recovery efforts. Customers reporting two similar failures did not rate firms as highly on recovery efforts as did those reporting distinct failures. These findings suggest a challenging implication: "Do not make the same mistake twice." Although it remains unlikely that firms will be able to avoid similar failures completely, managers can implement feedback loops into their service delivery system to reduce their occurrence.

Limitations and Research Issues

Although this study expands our knowledge of complaint handling, viable prospects for further research remain. Despite our evidence that noncomplainants are similar to complainants, it is possible that some customers chose not to complain about a failure but nonetheless expected a recovery. Although the bank encouraged complaints, it was the responsibility of customers to initiate a complaint. Therefore, further research could explore customer responses to proactive service recoveries initiated by the firm. It seems worthwhile to better understand if and how customers respond differently when firms proactively identify and successfully fix problems before customers complain (e.g., automobile recalls).

Although our results were mostly consistent across dependent variables, we found differences in univariate results between repurchase intent and WOM. For example, our double deviation data check after one failure and unsatisfactory recovery revealed different results for WOM and repurchase intent. In particular, whereas repurchase intent did not change from postfailure to postrecovery given one unsatisfactory recovery, WOM ratings increased. An unsatisfactory recovery following one failure had differential effects on types of intention, which suggests that even mildly unsatisfactory recoveries may spur increases in favorable WOM. Similarly, although post-Recovery 2 means for customers who reported a US sequence significantly increased above post-Failure 2 levels across the dependent variables collectively, this increase was not significant for repurchase intent. Perhaps complainants weigh past experiences more heavily when forming repurchase intent, which makes them less susceptible to recency effects. These results underscore the possibility that customers weigh and form various types of intentions differently. In a study of computer choice, Tsiros and Mittal (2000) find that satisfaction directly affects both purchase intent and complaint intent, but regret affects purchase intent only directly. Perhaps consumers used different processes to form complaint intentions and purchase intentions. Future studies can help clarify the cognitive and affective processes used to derive various behavioral intentions and help develop a greater understanding of the circumstances in which intentions remain stable or change over time.

Our research reinforces the notion that consumers' perceptions may change over time, signifying that perhaps what appears clear in cross-sectional studies may become complex in longitudinal studies--and vice versa. Our study joins a growing body of longitudinal research on consumer perceptions (e.g., Bolton and Lemon 1999; Mittal, Kumar, and Tsiros 1999), helping clarify and extend results found in cross-sectional studies. For example, Tax, Brown, and Chandrashekaran (1998) note that trust and commitment decrease when dissatisfaction with complaint handling increases. It seems fruitful to extend this finding by examining how trust and commitment change over time when customers report multiple failures with ongoing service providers. Similarly, extending the work by Smith, Bolton, and Wagner (1999), longitudinal studies could explore how the effects of service recovery attributes (e.g., response speed, apologies, compensation) on customer fairness perceptions change or remain stable when multiple failures occur.

Our study design reveals some potential measurement limitations that warrant examination. Although our prefailure retrospective measures of overall satisfaction were highly correlated with actual prefailure satisfaction ratings and there were no significant within-subject mean differences between our retrospective measures and actual measures, it still remains unclear when retrospective measures are accurate and when they are biased. For example, is there some time threshold (e.g., a certain number of months) within which customers can accurately recall their specific perceptions and after which their retrospective measures become biased? At what point do individual differences, environmental factors, customer involvement levels, and other factors spawn halo effects and other recall biases that cloud retrospective measures? To what extent do retrospective measures of given constructs trigger subsequent order effects when followed by a repeated measurement of the same constructs in the same questionnaire representing a different point in time (e.g., postfailure measures)? Given the challenges involved in capturing actual customer perceptions as they form over time, it seems worthwhile to investigate when retrospective measures offer reasonably accurate proxies.

Although all of our respondents received some type of redress effort in the bank's view, some or all of these efforts could have gone unnoticed or unappreciated by our respondents. A sound recovery in the bank's view may still be considered unsatisfactory or nonexistent in the customer's view. Alternatively, a customer could rate a recovery satisfactorily despite a lackluster recovery from the bank's view. As such, the same recovery effort from the bank's view could either generate paradoxical increases or spawn double deviations in customer ratings. Therefore, what one party considers a recovery may or may not be considered a recovery by the other party. Future work needs to examine if, when, and how customers and firm employees view recovery efforts differently.

Finally, the production and consumption of services are often inseparable, and customers may therefore influence the service they receive, including service recoveries. Relatively aggressive or passive customers, for example, may significantly affect the recovery process and ultimately influence their own perceptions about the experience. Does the "squeaky wheel get greased" or does the passive customer receive better recoveries? Although we captured the firm's response to complaints and how customers perceived these responses, we did not capture the extent to which customers influenced their recovery experiences. Although investigating the relationships between customer actions (as independent variables) and service experience evaluations was not the focus of this particular work, it offers a practical avenue for further research.

Notes:

1 From a between-subjects view, it is possible that satisfactory recoveries still spawn double deviations because of the dissatisfaction associated with a failure in the first place. Likewise, unsatisfactory recoveries can result in paradoxical increases just because customers are pleased that a firm at least tried to recover. Thus, the paradox and double deviation effect are not competing hypotheses.

• 2 We conducted a test of discriminant validity among constructs by comparing the average variance extracted estimates of all construct pairs with the phi correlation squared of the respective pairs (Fornell and Larcker 1981). We found discriminant validity across all pairs of constructs, time periods, and surveys. The phi correlations among constructs ranged from .83 (overall satisfaction with the firm and WOM of the first failure and recovery) and -.29 (failure severity and repurchase intent of the first failure and recovery). This information is available on request.
• 3 We also derived the unsatisfactory and satisfactory groups by conducting median splits and cluster analyses on the measures for satisfaction with recovery. For all analyses, these procedures produced results that closely resembled those of the midpoint split we employed.
• 4 We also ran the analysis by using a three-way split to create the time lag independent variable. All other aspects of our original model remained the same. We then compared the lower third to the upper third using linearly independent pairwise comparisons, and these results were relatively similar to our results using a median split. Furthermore, we also analyzed H12 and H13 through hierarchical regression. We modeled the months between failures as a continuous independent variable ranging from 1 to 20 months. The regression approach yielded the same conclusions as the RM MANCOVA approach employed to test H12, offering a multi-method reliability check of our analyses. The results are available on request.

Table 1: Linearly Independent Planned Comparisons: H1-H5

Legend for chart:
A - Dependent Variables
B - Per-Failure 2 Mean (SE)
C - Post-Recovery 2 Mean (SE)
D - Mean Difference
A
B                            C                D
H1: Two Satisfactory Service Recoveries (N = 74)
Satisfaction
15.49 (.476)            12.89 (.450)         -2.59**

Repurchase intent
22.16 (.536)            18.37 (.678)         -3.79**

WOM
15.06 (.424)            10.10 (.311)         -4.96**

H2 and H3: Two Unsatisfactory Service Recoveries (N = 76)
Legend for chart:
A - Dependent Variables
B - Post-Failure 2 Mean (SE)
C - Post-Recovery 2 Mean (SE)
D - H2: Mean Difference (j, T4-T3b)
E - Mean Difference (i, T2-T1b)
F - H3: Mean Difference (j - i)
A
B                C                 D            E           F
Satisfaction
6.09 (.350)     2.93 (.461)      -3.16**       .80        -3.96**

Repurchase intent
13.17 (.599)    5.23 (.696)      -7.94**      -.20        -7.74**

WOM
5.06 (.290)     3.16 (.319)      -1.90**       2.13       -4.03**

US Recovery Sequence (N = 69)
Legend for chart:
A - Dependent Variables
B - Post-Failure 2 Mean (SE)
C - Post-Recovery 2 Mean (SE)
D - H5: Mean Difference
A
B                          C                    D
Satisfaction
5.85 (.475)            10.82 (.627)            4.97**

Repurchase intent
12.90 (.814)           14.57 (.945)            1.67

WOM
5.61 (.394)             9.36 (.433)            3.75**

SU Recovery Sequence (N = 36)
Legend for chart:
A - Dependent Variables
B - Post-Failure 2 Mean (SE)
C - Post-Recovery 2 Mean (SE)
D - H4: Mean Difference
E - H4: Post-Recovery 2 Mean Difference (US-SU)
A
B                C                   D               E
Satisfaction
6.33 (.364)      5.03 (.479)        -1.30*         5.79**

Repurchase intent
13.64 (.673)     8.88 (.723)        -4.76**        5.35**

WOM
5.83 (.301)      4.01 (.331)        -1.82**        5.69**

*p < .05.
**p < .01.

Notes: Estimated marginal means reported are adjusted for the effects of failure severity, attributions of blame, and recovery expectations. All variables are based on summed-item scores. SE = standardized error, which is reported for estimated marginal means.

Table 2: Linearly Independent Planned Comparisons: H6-H11

Legend for chart:
A - All Respondents (N=255)
B - Failure 1 Mean (SE)
C - Failure 2 Mean (SE)
D - Mean Difference

Recovery expectations
16.90(.370)        20.46(.354)        3.56**

Failure severity
12.81(.323)        15.21(.303)        2.40**

Attributions of blame
14.96(.270)        17.43(.246)        2.47**

Group: Unsatisfactory Service
Recovery (N=112)
Recovery expectations
18.59(.532)        21.14(.510)        2.55**

Failure severity
14.41(.465)        15.76(.436)        1.35*

Attributions of blame
13.71(.389)        17.64(.354)        3.92**

Group: Satisfactory Service
Recovery (N=143)
Recovery expectations
15.21(.513)        19.77(.492)        4.56**

Failure severity
11.21(.448)        14.65(.420)        3.45**

Attributions of blame
16.20(.375)        17.23(.341)        1.03*


                             Univariate Statistics
Model                                  F         Effect size (η(2))
H6: Expectations X failure            49.84           .17**

H7: Expectations X failure X recovery  3.97           .02*

H8: Severity X failure                30.40           .11**

H9: Severity X failure X recovery      5.78           .02*

H10: Attributions X failure            47.47           .16**

H11: Attributions X failure X recovery 16.22           .06**

(*)p<.05 (**)p<.01. Notes: Based on estimated marginal means, controlling for the effect of failure type. All variables are based on summed-item scores. SE= standardized error.

Table 3: Linearly Independent Planned Comparisons: H12-H15

Legend for chart:
A - H12
B - Group: Shorter Gap Between Failures (Mean=2.61, N=128)
C - Group: Longer Gap Between Failures (Mean=10.69, N=127)
D - Mean Difference
A
B                       C               D
Post-Failure 2 Means (SE)
Satisfaction
6.11 (.280)         6.24 (.281)        .13

Repurchase intent
12.94 (.418)       13.35 (.419)        .41

WOM
6.01 (.294)         5.31 (.295)       -.70

A - H13
Post-Recovery 2 Means (SE)
Satisfaction
3.96 (.356)        11.07 (.358)       7.10*

Repurchase intent
6.52 (.490)        16.22 (.492)       9.70*

WOM
3.34 (.222)         9.24 (.223)       5.90*

Post-Failure 2 Means (SE)
Legend for chart:
A  - H14
B - Group: Similar Failures (N=118)
C - Group: Different Failures (N=137)
D - Mean Difference
Satisfaction
6.05 (.299)        6.42 (.284)         .38

Repurchase intent
12.65 (.444)       13.78 (.422)       1.13

WOM
5.47 (.316)        5.86 (.300)        .39

Post-Recovery 2 Means (SE)
A = H15

Satisfaction
6.50 (.434)        8.05 (.412)         1.55*

Repurchase intent
9.34 (.600)        12.74 (.569)        3.40*

WOM
5.16 (.262)         6.78 (.249)        1.63*

*p < .01. Note: H12 and H13 were based on estimated marginal means, controlling for the effect of post-Recovery 1 satisfaction. H14 and H15 were based on estimated marginal means, controlling for the effects of post-Recovery 1 satisfaction and failure type. All variables are based on summed-item scores. SE = standardized error.

Figure 1: Time Line of Measurement

Bolton, Ruth N. and James H. Drew (1991), "A Multistage Model of Customers' Assessments of Service Quality and Value," Journal of Consumer Research, 17 (March), 375-84.

------ and Katherine N. Lemon (1999), "A Dynamic Model of Customers' Usage of Services: Usage as an Antecedent and Consequence of Satisfaction," Journal of Marketing Research, 36 (May), 171-86.

Boulding, William, Ajay Kalra, Richard Staelin, and Valarie A.

Zeithaml (1993), "A Dynamic Process Model of Service Quality: From Expectations to Behavioral Intentions," Journal of Marketing Research, 30 (February), 7-27.

Cronin, J. Joseph, Jr., and Steven A. Taylor (1994), "SERVPERF Versus SERVQUAL: Reconciling Performance-Based and Perceptions-Minus-Expectations Measurement of Service Quality," Journal of Marketing, 58 (January), 125-31.

Folkes, Valerie S. (1984), "Consumer Reactions to Product Failure: An Attributional Approach," Journal of Consumer Research, 10 (March), 398-409.

------ (1988), "Recent Attribution Research in Consumer Behavior: A Review and New Directions," Journal of Consumer Research, 14 (March), 548-65.

------ and Barbara Kotsos (1986), "Buyers' and Sellers' Explanations for Product Failure: Who Done It?" Journal of Marketing, 50 (April), 74-80.

Fornell, Claes and David F. Larcker (l981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Errors," Journal of Marketing Research, 18 (February), 39-50.

Gilly, Mary C. and Betsy D. Gelb (1982), "Post-purchase Consumer Processes and the Complaining Consumer," Journal of Consumer Research, 9 (December), 323-28.

Grayson, Kent and Tina Ambler (1999), "The Dark Side of Long-Term Relationships in Marketing Services," Journal of Marketing Research, 36 (January), 132-41.

Hart, Christopher W.L., James L. Heskett, and W. Earl Sasser Jr. (1990), "The Profitable Art of Service Recovery," Harvard Business Review, 68 (July/August), 148-57.

Hornik, Jacob (1984), "Subjective vs. Objective Time Measures: A

Note on the Perception of Time in Consumer Behavior," Journal of Consumer Research, 11 (June), 615-18.

Kahneman, Daniel and Amos Tversky (1979), "Prospect Theory: An Analysis of Decision Under Risk," Econometrica, 47 (March), 263-91.

LaBarbera, P.A. and D. Mazursky (1983), "A Longitudinal Assessment of Consumer Satisfaction/Dissatisfaction: The Dynamic Aspect of the Cognitive Process," Journal of Marketing Research, 9 (December), 323-28.

Loewenstein, George F. and Drazen Prelec (1993), "Preferences for Sequences of Outcomes," Psychological Review, 100 (1), 91-108.

McCollough, Michael A., Leonard L. Berry, and Manjit S. Yadav (2000), "An Empirical Investigation of Customer Satisfaction After Service Failure and Recovery," Journal of Service Research, 3 (November), 121-37.

Mittal, Vikas, Pankaj Kumar, and Michael Tsiros (1999), "Attribute-Level Performance, Satisfaction, and Behavioral Intentions over Time: A Consumption-System Approach," Journal of Marketing, 63 (April), 88-101.

------, William T. Ross Jr., and Patrick M. Baldasare (1998), "The Asymmetric Impact of Negative and Positive Attribute-Level Performance on Overall Satisfaction and Repurchase Intentions," Journal of Marketing, 62 (January), 33-47.

Neter, John, Michael H. Kutner, Christopher J. Nachtsheim, and William Wasserman (1996), Applied Linear Statistical Models. Chicago: Richard D. Irwin.

Oliver, Richard L. (1980), "A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions," Journal of Marketing Research, 17 (November), 460-69.

------ (1997), Satisfaction: A Behavioral Perspective on the Consumer. Boston: McGraw-Hill/Irwin.

------ and Raymond R. Burke (1999), "Expectation Processes in Satisfaction Formation: A Field Study," Journal of Service Research, 1 (3), 196-214.

Ross, William T., Jr., and Itamar Simonson (1991), "Evaluations of

Pairs of Experiences: A Preference for Happy Endings," Journal of Behavioral Decision Making, 4 (4), 273-82.

Seiders, Kathleen and Leonard L. Berry (1998), "Service Fairness: What It Is and Why It Matters," Academy of Management Executive, 12 (2), 8-20.

Smith, Amy K. and Ruth N. Bolton (1998), "An Experimental Investigation of Customer Reactions to Service Failure and Recovery Encounters: Paradox or Peril," Journal of Service Research, 1 (1), 5-17.

------, ------, and Janet Wagner (1999), "A Model of Customer Satisfaction with Service Encounters Involving Failure and Recovery," Journal of Marketing Research, 36 (August), 356-73.

Tax, Stephen S. and Stephen W. Brown (1998), "Recovering and

Learning from Service Failures," Sloan Management Review, 40 (Fall), 75-89.

------, ------, and Murali Chandrashekaran (1998), "Customer Evaluations of Service Complaint Experiences: Implications for Relationship Marketing," Journal of Marketing, 62(April),60-76.

Tsiros, Michael and Vikas Mittal (2000), "Regret: A Model of Its Antecedents and Consequences in Consumer Decision Making," Journal of Consumer Research, 26 (March), 401-17.

Vonesh, Edward and Vernon Chinchilli (1997), Linear and Nonlinear Models for the Analysis of Repeated Measurements. New York: Marcel Dekker.

Weiner, Bernard (2000), "Attributional Thoughts About Consumer Behavior," Journal of Consumer Research, 27 (December), 382-87.

Measurement Scales

Overall Firm Satisfaction[a]

1. I am satisfied with my overall experience with [firm name].[b]

• 2. As a whole, I am not satisfied with [firm name].
• 3. How satisfied are you overall with the quality of [firm name] banking service?[b]

Favorable WOM[a]

1. How likely are you to spread positive word-of-mouth about [firm name]?

• 2. I would recommend [firm name's] banking services to my friends.
• 3. If my friends were looking for a banking service, I would tell them to try [firm name].

Repurchase Intent[a]

1. In the future, I intend to use banking services from [firm name].

• 2. If you were in the market for additional banking services, how likely would you be to use those services from [firm name]?
• 3. In the near future, I will not use [firm name] as my provider.
• 4. In the future, I will continue using [firm name] for these banking services.

Service Recovery Expectations[c]

1. I have high expectations that [firm name] will fix the problem.

• 2. I expect [firm name] to do whatever it takes to guarantee my satisfaction.
• 3. I think [firm name] will quickly respond to (banking) problems.
• 4. My expectations are high that I will receive compensation when I encounter a banking service problem.

Failure Severity[c]

In my opinion, the banking problem that I experienced was a

1. Minor problem ( 1)/major problem ( 7).

• 2. Big inconvenience ( 1)/small inconvenience ( 7).
• 3. Major aggravation ( 1)/minor aggravation ( 7).

Attributions of Blame[c]

1. To what extent was [firm name] responsible for the problem that you experienced? (not at all responsible

[ 1]/totally responsible [ 7])

• 2. The problem that I encountered was all [firm name]'s fault.
• 3. To what extent do you blame [firm name] for this problem? (not at all [ 1]/completely [ 7])

Satisfaction with Service Recovery[d]

1. In my opinion, [firm name] provided a satisfactory resolution to my banking problem on this particular occasion.

• 2. I am not satisfied with [firm name]'s handling of this particular problem.
• 3. Regarding this particular event (most recent banking problem), I am satisfied with [firm name].

[a]Measured at all time periods.

[b]Indicates that the scale was anchored with "not at all satisfied" and "very satisfied."

[c]Measured once at T1b (post-Failure 1) and again at T3b (post-Failure 2).

[d]Measured once at T2 (post-Recovery 1) and again at T4 (post-Recovery 2). Notes: All items were measured on a seven-point scale. Unless noted, all items were anchored with "strongly disagree" and "strongly agree."

Construct Raw Means, Standard Deviations, and Pearson Correlations

Legend for chart:
A - Construct
B - Mean
C - S.D.
D - 1
E - 2
F - 3
G - 4
H - 5
I - 6
J - 7
K - 8
L - 9
M - 10
N - 11
O - 12
P - 13
Q - 14
R - 16
S - 17
T - 18
U - 19
V - 20
W - 21
X - 22
Y - 23
Z - 24

A
B       C      D      E      F     G     H     I      J     K     L
M       N      O      P      Q     R     S     T      U     V     W
X       Y      Z


1. Pre-Failure 1 SAT, F1
13.38   2.95   .87

2. Pre-Failure 1 RI, F1
22.16  5.59    .55    .90

3. Pre-Failure 1 WOM, F1
9.49   3.19    .57   -.08   .83

4. Post-Failure 1 SAT, F1
8.89   3.93    .24   -.31   .75   .87

5. Post-Failure 1 RI, F1
16.31  3.74    .51    .62   .31   .15   .83

6. Post-Failure 1 WOM, F1
7.90   3.68    .09   -.34   .6    .63   .22    .86

7. Post-Recovery I SAT, F1
13.04  5.08    .39    .23  .02   -.06   .18   -.31   .93

8. Post-Recovery 1 RI, F1
19.58  6.67    .24    .37  -.05   .02   .45   -.23   .72   .96

9. Post-Recovery l WOM, F1
12.25  5.15   -.03   -.24   .02   .11  -.06   -.12   .62   .49   .93

10. Attributions of blame, F1
15.19  4.25    .26    .48  -.36  -.49   .13   -.61   .51   .40   .28
 .87

11. Failure severity, F1
12.69  5.05   .21    .28   -.28  -.60  -.05   -.26  -.04   -.26  -.24
  .51   .88

12. Recovery expectations, F1
16.89  5.77   .36   .18    -.01  -.42   .02   -.30   .12   -.19  -.08
  .45   .69   .92

13. Pre-Failure 2 SAT, F2
12.76  5.20   .02  -.02    -.05   .03   .10   -.02   .50     .48  .48
  .16  -.23  -.17   .92

14. Pre-Failure2 RI, F2
19.34  6.82   .0   -.10    -.05   .04   .03   -.06   .50     .43  .52
  .22  -.18  -.14   .73     .89


15. Pre-Failure 2 WOM, F2
12.33  5.12  -.09  -.10    -.12   .06   .03    -.01  .51     .51  .58
  .19  -.29  -.28   .63     .53   .89

16. Post-Failure 2 SAT, F2
6.18   3.12  -.04   .03    -.08   .00   .03     .05  .06     .12  .12
 .01   -.08  -.13   .15    -.10   .51    84

17. Post-Failure 2 RI, F2
13.14  4.66   .01   .00    -.03   .02   .03    -.04  .05     .08  .07
  .01  -.06  -.02   .41     .53   .30   .27     .86


18. Post-Failure 2 WOM, F2
5.66  3.30  -.05    .02    -.03   .01   .08     .07   .04    .11  .09
 .01  -.06  -.11    .02    -.23   .42   .69     .09   .87

19. Post-Recovery 2 SAT, F2
7.50  5.37   .03   -.05     .09   .14   .07     .11   .15    .16  .17
 .04  -.01  -.08    .33     .35   .22  -.07     .30  -.15    .97


20. Post-Recovery 2 RI, F2
11.35  7.50  .01   -.08     .03   .13   .05     .08   .20    .24  .27
  .05  -.08 -.15    .34     .45   .27  -.01     .41  -.11    .81  .97

21. Post-Recovery 2 WOM, F2
6.28  3.88   .03  -.03      .03   .10   .03     .05   .20    .22  .13
 .02  -.07  -.10   .11     -.01   .10  -.02    -.19  -.08    .57  .51
 .96

22. Attributions of blame, F2
17.43 3.68  .08    .01     .02   -.01   .02    -.07   .03  -.02  -.05
  .00  .03  .14    .26     .39   -.22  -.51     .21  -.75   .29   .24
  .05  .88


23. Failure severity, F2
15.03  4.61 -.02   .00     .01   -.07   .02    -.24  -.12  -.16  -.14
 -.04   .09  .08   .05     .09   -.21  -.42    -.05  -.23   .05  -.13
 -.31   .42  .89

24. Recovery expectations, F2
20.31  5.34  -.01 -.02     .02    .02   .04     .01  -.10  -.12  -.11
 -.10   .00   .08  .20     .14   -.13  -.39     .12  -.29   .19   .00
 -.21   .47   .72  .92

Notes: F1= Failure 1 and F2= Failure 2. SAT= overall satisfaction; RI= repurchase intentions. Coefficient alphas are reported on the diagonal.

~~~~~~~~

By James G. Maxham III and Richard G. Netemeyer

James G. Maxham III is Assistant Professor of Commerce, and Richard G. Netemeyer is Professor of Commerce, McIntire School of Commerce, University of Virginia. This research was funded in part by the Bernard A. Morin Fund for Marketing Excellence at the McIntire School of Commerce. The authors thank Amanda Bower, David Mick, Bill Bearden, and the University of Virginia Statistics department for contributing insightful comments on previous versions of this article.

A Marketing Perspective on Mergers and Acquisitions: How Marketing Integration Affects Postmerger Performance 

Previous research on mergers and acquisitions (M&A) has neglected marketing issues by and large. In this article, the authors examine the effects of postmerger integration in marketing (extent and speed of marketing integration) on M&A performance, as mediated by integration outcomes (magnitude of cost savings and market-related performance). Results from a survey of 232 horizontal M&A show that market-related performance after the merger or acquisition has a much stronger impact on financial performance than does cost savings. In addition, the authors find that the extent of integration is beneficial in terms of cost savings but detrimental in terms of market-related performance. Finally, they identify variables that moderate the relationships being considered.

Mergers and acquisitions (M&A) have become increasingly popular in business practice, and especially horizontal M&A (M&A that take place in the same industry, often between direct competitors) are undertaken frequently (Krishnan and Park 2002). However, there is considerable evidence that many M&A activities are unsuccessful. Estimated failure rates are typically between 60 and 80% (Marks and Mirvis 2001; Tetenbaum 1999). Thus, studying factors that influence the success of M&A is a promising field for academic research.

Because there is a growing recognition that "all value creation takes place after the acquisition" (Haspeslagh and Jemison 1991, p. 129), the topic of postmerger integration (PMI) has received increasing attention (e.g., Capron, Dussauge, and Mitchell 1998; Datta 1991; Larsson and Finkelstein 1999; Shrivastava 1986). However, marketing-related issues of PMI, such as whether and how two firms' marketing activities are integrated and how this affects the merged firms performance, have been largely neglected.

Within the marketing discipline, M&A-related research is almost totally absent. A notable exception is a study by Capron and Hulland (1999). Their findings indicate that the redeployment of marketing resources has a significant effect on firm performance after M&A.

The lack of attention given to marketing issues in the context of M&A is in sharp contrast with many statements that highlight the importance of marketing-related issues for M&A performance (e.g., Becker and Flamer 1997; Clemente and Greenspan 1997). For example, Bekier and Shelton (2002) report that there is considerable risk of losing customers in M&A. During the integration phase, managerial energy is often absorbed by internal issues, which can lead managers to neglect customer-related tasks (Hitt, Hoskisson, and Ireland 1990). This strong internal orientation is frequently accompanied by a decline in service quality (Urban and Pratt 2000). On the part of customers, this decline can result in uncertainties about their future relationship with the merging firms (e.g., prices, quality of products and services, contact persons). Restraint and defection are possible customer reactions (Chakrabarti 1990; Reichheld and Henske 1991). Competitors' actions often reinforce this effect as they take advantage of the situation and try to alienate customers. The relevance of market-related issues for the success of a PMI is also highlighted by Morall (1996, p. 19), who claims on the basis of a case study that "cost reduction to make a merger pay off is not as important as customer retention." Because such market-related issues have been neglected in previous research, we argue that the research-based understanding of the factors that lead to M&A success is still limited to date.

Against this background, we adopt a marketing perspective of M&A and explore three issues. First, we investigate how the marketing integration process (extent and speed of integration) affects integration outcomes (magnitude of cost savings and market-related performance). Second, we investigate how these relationships are affected by certain moderators (customer orientation of integration, market growth, product/service industry, relatedness of the firms market positioning, and relative size of the acquired firm). Third, we analyze the importance of market-related performance (compared with cost savings) for M&A performance.

Theoretical Background

Our study is rooted in the resource-based theory, which explains superior performance by specific resources owned by the firm (Barney 1991; Peteraf 1993; Wernerfelt 1984). Following Barney (1991, p. 101), we define resources as assets, organizational processes, capabilities, firm attributes, information, knowledge, etc. controlled by a firm that enable the firm to conceive of and implement strategies that improve its efficiency and effectiveness. According to this theory, resources must meet four requirements--value, rarity, imperfect imitability, and insubstitutability (Barney 1991)--to affect firm performance.

The resource-based view has been used frequently as a theoretical explanation for why M&A occur (e.g., Anand and Singh 1997; Karim and Mitchell 2000). Many organizational theorists argue that firms face constraints (in terms of intellectual capabilities and time) with respect to the internal development of new resources (e.g., Nelson and Winter 1982; Singh and Montgomery 1987). Therefore, many firms turn to the market to obtain new resources (Capron, Dussauge, and Mitchell 1998). Because few resources are available as single entities, firms must acquire entire businesses to extract value from the resources owned by the acquired firm (Barney 1986).

A key issue in M&A research that adopts a resource-based perspective is to what extent the merging firms integrate certain resources after a merger or acquisition to achieve a stronger competitive position and thus superior financial performance. As an example, Capron, Dussauge, and Mitchell (1998) analyze performance implications associated with the integration of research and development; manufacturing; marketing; and managerial, financial, and senior executive resources. These authors argue that among the resources with the potential to contribute to firms postmerger performance, marketing resources, such as brands and sales forces, are a highly important subset. In addition, marketing resources have been identified as resources that meet the previously mentioned requirements for relevancy for firm performance (Capron and Hulland 1999).

Thus, the resource-based perspective provides a theoretical basis for the fundamental proposition of our study, namely, that the integration of marketing resources is relevant for M&A performance. We translate this proposition into specific constructs and hypotheses.

Framework

Our unit of analysis is the total marketing activities of two merged firms. The framework of our study is a causal chain with three categories of constructs: marketing integration process, integration outcomes, and performance outcome after the merger or acquisition. Figure 1 presents an overview of the framework and the specific constructs.

Marketing integration process We define the extent of integration as the level of similarity achieved between two firms marketing systems, structures, activities, and processes. As an example, a typical question in PMI pertains to what extent the differences between two firms, in terms of their product and service offer, branding strategy, sales channels, and so forth, should be consolidated. A high level of integration means that premerger differences between the two organizations have largely disappeared. This integration can be achieved by using a system, structure, activity, or process of either company for both firms. It also can be achieved through combining the "best of both worlds" into a newly created system, structure, activity, or process. A maximal extent of integration is achieved if all differences in the firms marketing approaches are harmonized. The extent of integration has been discussed as a key driver of M&A performance (e.g., Birkinshaw, Bresman, and Hakanson 2000).( n1)

Speed of integration is defined as the shortness of the time needed to achieve the intended level of marketing integration. Although the importance of integration speed has been highlighted in many practitioner statements (e.g., Bragado 1992; Mitchell 1989), this construct has been almost totally neglected in academic research (with the exception of some qualitative work in human resources management; e.g., Buono and Bowditch 1989; Schweiger and Goulet 2000).

Integration and performance outcomes. A distinction between cost-related and market-related integration outcomes is common (Chatterjee 1986). We define magnitude of cost savings as the extent to which the merging firms reduce the amount of resources and thus costs (in terms of personnel, infrastructure, and so on) compared with the sum of their resources before the transaction. The magnitude of cost savings is minimal if the postintegration costs are identical to the sum of the two firms' premerger costs.

In addition to gains from cost reductions, M&A performance can also be influenced by the merging firms ability to increase market-related performance and thus revenues after the transaction. We define market-related performance after the merger or acquisition as the effectiveness of the combined firms' marketing activities. Positive market-related outcomes of M&A (e.g., customer loyalty, market share) may be based on phenomena such as intensified cross-selling and the use of expanded bundling opportunities (because of a broader product range), as well as improved negotiating positions toward customers.

The key performance measure in our study is financial performance after the merger or acquisition. Consistent with existing literature, we define the financial performance after the merger or acquisition as the postmerger profitability compared with the firms' premerger profitabilities (Datta 1991; Hunt 1990). This after/before comparison is typically used as an indicator of M&A success or failure (e.g., Capron and Hulland 1999).

Moderator variables. Potential moderating influences arise at the firm level and the market/industry level (see Figure 1). We based the selection of specific moderator variables on two considerations. First, previous research on M&A has identified two constructs (i.e., relatedness of the firms market positioning and relative size of the acquired firm) with potential relevance for our study. Second, we identified moderator variables (i.e., customer orientation, market growth, and product/service industry) on the basis of previous marketing research. We describe these moderator variables and justify their relevance in more detail in the following section.

Main Effects

Our first hypothesis addresses the link between the extent of integration and the magnitude of cost savings. The basic logic is that a merger or acquisition creates redundancies that can be eliminated through integration. More specifically, integration activities create the potential for a reduction in the amount of needed resources and thus costs (e.g., personnel, infrastructure) (Capron 1999; Seth 1990a). Hypothesizing a positive impact of the extent of integration on cost savings is consistent with previous findings by Capron (1999). Thus,( n2)

H₁: The extent of integration is positively associated with the magnitude of cost savings.

Although we hypothesize that integration is beneficial in terms of cost savings, it is likely to be detrimental in terms of market-related performance. Extensive marketing integration, in practical terms, leads to a reduced number of brands, product variants, service offers, distribution channels, and so forth. Therefore, the merged firms' ability to adapt their offer to the needs of specific market segments is reduced through extensive integration. This reduction can be detrimental for market-related performance because some customers may perceive the value provided by the merged firm as less attractive. In addition, in the case of extensive integration, firms are likely to reassess their joint customer portfolio and eliminate unprofitable customers or segments (i.e., "controlled" loss of customers and market share) (Reichheld and Henske 1991).

Furthermore, during PMI, managerial energy is often strongly absorbed by internal issues at the expense of customer-related tasks (Hitt, Hoskisson, and Ireland 1990). This absorption is true for a high extent of integration because, in such circumstances, many internal decisions (e.g., structures, staffing) must be prepared and made. Therefore, in the case of a high extent of integration, the neglect of market-and customer-related tasks is more likely than in the case of a low extent of integration. Customer dissatisfaction, restraint, and defection are likely consequences. Thus,( n3)

H₂: The extent of integration is negatively associated with market-related performance after the merger or acquisition.

Our reasoning regarding integration speed is based on the logic that an M&A activity creates uncertainty among customers, which may ultimately lead to customer defection. Uncertainty is likely to increase over time if the integration phase is long. We argue that the major beneficial effect of integration speed on M&A success is the uncertainty reduction for customers. If integration decisions are made and implemented quickly, customers will know what to expect from the merged company in terms of the product offer, pricing policy, contact persons, and so forth. Thus, customer uncertainty is reduced through a high level of integration speed. In addition, it is a common business practice for competitors to try to increase uncertainty among the merging firms' customers to promote customer switching (Clemente and Greenspan 1997). This potential source of customer uncertainty can also be eliminated through high integration speed.

Finally, a fast integration will lead to a reduced level of uncertainty among employees about their future in the merged company. In particular, the uncertainty of customer contact personnel about their future in the company can lead to increased customer uncertainty and thus be harmful for market-related performance (e.g., Singh, Goolsby, and Rhoads 1994). Thus,

H₃: The speed of integration is positively associated with market-related performance after the merger or acquisition.

We further argue that market-related performance has a positive impact on the magnitude of cost savings for two reasons. First, a high level of customer loyalty after the transaction (which is a facet of market-related performance) enables companies to avoid costly activities directed toward attracting new customers. Retaining customers is much less cost intensive than is attracting new customers (e.g., Anderson, Fornell, and Rust 1997). Second, if market share (and consequently sales volume) increases, opportunities for cost savings through economies of scale emerge. Our reasoning is also supported by findings that indicate a negative impact of market performance on costs (Phillips, Chang, and Buzzell' 1983). Thus,

H₄: The market-related performance after the merger or acquisition is positively associated with the magnitude of cost savings.

We finally suggest that the magnitude of cost savings and market-related performance both lead to superior postmerger financial performance. Because financial performance is defined in terms of profitability, the cost-related effect is obvious: If costs decrease, all other things being equal, profitability increases. Market-related performance after the merger or acquisition affects profitability through increased sales. Positive relationships between market share and profitability (Szymanski, Bharadwaj, and Varadarajan 1993) and between customer loyalty and profitability (Kalwani and Narayandas 1995) have been documented in previous research. In addition, hypothesizing these two positive effects is in line with previous empirical findings in M&A research (e.g., Anand and Singh 1997). This leads us to the following hypotheses:

H₅: The magnitude of cost savings is positively associated with the financial performance after the merger or acquisition.

H₆: The market-related performance after the merger or acquisition is positively associated with the financial performance after the merger or acquisition.

Moderating Effects

We develop hypotheses regarding the firm-level moderators (customer orientation of integration, relatedness of the firms' market positioning, relative size of the acquired firm). Subsequently, we discuss the effects of the market-and industry-level moderators (market growth before the merger or acquisition, distinction between product and service firms).

Customer orientation of integration. As we described previously, PMI is often characterized by the strongly internal orientation of managers (Hitt, Hoskisson, and Ireland 1990). A possible consequence is that decisions are made predominantly on the basis of internal criteria, such as internal structures, processes, power distribution, or individual managers' preferences. Against this background, we define customer orientation of integration as the extent to which decisions about marketing integration are driven by customer-related considerations rather than internal considerations. For example, a high level of customer orientation of integration is present if decisions are strongly influenced by the goal of creating additional customer value rather than reducing the costs of serving customers.

In the case of a high customer orientation in PMI, decisions are driven by cost reduction motives to a lesser extent.

Therefore, an increase in the extent of integration will produce less cost savings in this situation than in a context of low customer orientation. Furthermore, we predict that a high level of customer orientation can alleviate, at least to some extent, the negative market-related consequences of integration. When decisions to integrate brands, product variants, or distribution channels are made with a strong focus on customer value, the negative market-related consequences of a high level of the extent of integration should be weaker.

Our argument regarding integration speed is that speed serves as a means to reduce uncertainty among customers. We now argue that customer orientation serves as a partial substitute for speed in reducing customer uncertainty. More specifically, when customers perceive that integration decisions are driven by customer-related considerations, their trust in the newly forming firm will increase and reduce uncertainty. In other words, in the case of a high customer orientation, integration speed becomes less relevant as a means to reduce customer uncertainty and avoid detrimental effects on market-related performance. We therefore predict that the effect of integration speed on market-related performance is weaker in the case of high rather than low customer orientation. In summary, we hypothesize the following:

H₇: In the case of high rather than low customer orientation of integration, the effect of the (a) extent of integration on cost savings is less positive, (b) extent of integration on market-related performance after the merger or acquisition is less negative, and (c) speed of integration on market-related performance after the merger or acquisition is less positive.

Relatedness of the firms' market positioning. We define relatedness of the firms' market positioning as the extent to which the firms offers are similar in terms of the customer needs that they satisfy, quality, and price positioning. This variable has been frequently analyzed in M&A research (e.g., Chakrabarti 1990). Unlike previous research, our research question is not related to the possible direct performance effect of relatedness but rather to how relatedness affects the way that PMI should be conducted.

Consistent with previous studies (e.g., Hagedorn and Duysters 2002), we argue that a high level of relatedness offers a great potential for cost reductions. Thus, an increase in the extent of integration may result in greater cost savings when the market positioning of the firms is highly related. However, we do not find a compelling argument for why the relatedness of market positioning should moderate the impact of the extent of integration on market-related performance.

The potential scope of changes (e.g., repositioning the strategic focus of the entire firm) is much greater in M&A between unrelated firms (Larsson 1989). Because of the reduced potential for changes in highly related M&A, a lower level of uncertainty among customers about their future relationship with the merged firm is likely. Because uncertainty reduction is the major effect of integration speed, we predict that an increase of integration speed has a smaller impact on market-related performance when relatedness is high. Thus, we hypothesize the following:

H₈: In the case of high as opposed to low relatedness between the merging firms' market positioning, the effect of the (a) extent of integration on cost savings is more positive and (b) speed of integration on market-related performance after the merger or acquisition is less positive.

Relative size of the acquired firm. We define the relative size of the acquired firm as the relative annual sales volume of the acquiree compared with that of the acquirer. This variable has been frequently analyzed in previous M&A research. There is substantial evidence that the potential to create value from M&A depends on relative size (e.g., Capron 1999; Seth 1990b).

In the case of high relative size (i.e., when the acquiree is almost as big as the acquirer), there is a greater increase in scale than in the case of low relative size. Therefore, there is a greater potential for cost savings through integration in this case. In addition, because organizational size is known to be an important driver of organizational structure (e.g., Blau and Schoenherr 1971; Pugh et al. 1969), we argue that there is more structural similarity between the organizations in the case of high relative size. For example, firms with similar size should be structurally similar with respect to the presence or nonpresence of certain specialized units (e.g., market research, key account management) within the company. In other words, there will be a larger set of similar structures and therefore more redundancies in the case of high relative size so that the potential for cost reduction is greater.

With respect to market-related performance, we predict that the negative effect of the extent of integration (H₂) is stronger in the case of a high relative size. If the acquiree is fairly large compared with the acquirer, PMI is likely to involve more people, and its results will not be as clear from the beginning compared with the situation in which a relatively small firm has been acquired. In these conditions, internal conflicts, interorganizational competition, holding back of information, and so forth, are likely consequences and will absorb managerial energy that is needed to serve customers. Against this background, an increase in the integration extent will produce greater damage to market-related performance when the relative size of the acquired firm is high.

With respect to integration speed, our reasoning related to relative size as a moderator variable is based on the contention that the number of customers affected by the transaction is greater when the relative size is high. When the acquired firm is relatively small in proportion to the acquirer, only a small number of customers are affected by the integration activities. With an increasing relative size, the customer base affected by the integration also broadens. In turn, the potential for rumors about possible changes, which lead to uncertainties among customers, also increases. Thus, an increase of integration speed is more likely to be beneficial for market-related performance when the relative size of the acquiree is high. Thus,

H₉: In the case of high rather than low relative size of the acquired firm, the effect of the (a) extent of integration on cost savings is more positive, (b) extent of integration on market-related performance after the merger or acquisition is more negative, and (c) speed of integration on market-related performance after the merger or acquisition is more positive.

Market growth. Market growth is a frequently studied construct in marketing research. We consider market growth before the merger or acquisition and argue that the potential for cost savings through extensive integration is greater in mature markets (i.e., markets with low growth rates) than in high growth markets (e.g., Budros 1999). This reasoning is based on the proposition that in mature markets, companies have accumulated a higher level of experience with respect to business processes and routines (Davis and Thomas 1993), which facilitates cost savings. Thus, extensive integration in fast growing markets results in smaller cost savings than in low growth markets.

We also hypothesize a moderating effect of market growth on the link between the extent of integration and market-related performance. The logic behind our hypothesis is that in markets with low growth rates (which are typically in later life cycle stages), there is less dynamism than in high growth markets (e.g., Wasson 1978). For example, market shares (e.g., Arndt 1979) and the competitive environment (e.g., less entry of new competitors; Day 1981) are more stable, and switching barriers, which make it costly and risky for a customer to change its supplier, are typically higher in mature markets (e.g., Colgate and Lang 2001; Fornell 1992). Switching barriers may limit customer defection and thus reduce negative customer reactions to a high extent of integration. Moreover, the number of competitors is usually lower in low growth than in high growth markets (e.g., Buzzell 1981). Therefore, there are fewer alternative suppliers for the customer, and fewer competitors' actions to alienate customers will occur in low growth markets. As a consequence, the negative impact of the extent of integration on market-related performance is stronger in high growth than in low growth markets.

Our reasoning with respect to the impact of integration speed on market-related performance is based on a similar logic. Because customers in mature markets are less likely to defect because of high switching barriers, fewer alternative suppliers, or fewer competitors' actions, we predict that the positive effect of speed is weaker in low growth markets. This hypothesis is consistent with research that suggests that speed is less important if markets are growing slowly (Bowman and Gatignon 1995). In summary, we present the following hypothesis:

H₁₀: In the case of high rather than low market growth before the merger or acquisition, the effect of the (a) extent of integration on cost savings is less positive, (b) extent of integration on market-related performance after the merger or acquisition is more negative, and (c) speed of integration on market-related performance after the merger or acquisition is more positive.

Manufacturing versus service firms. Services exhibit higher levels of individualization than tangible products (e.g., Zeithaml, Parasuraman, and Berry 1985). Therefore, we predict that translating a high extent of integration into significant cost savings is more difficult for firms that market services than for firms that market tangible products. In other words, we predict that the effect of the extent of integration on cost savings is stronger for product than for service firms. However, we do not find a compelling argument for why the distinction between product and service firms should moderate the impact of the extent of integration on market-related performance.

In addition, a higher level of uncertainty among customers is a key characteristic of services (Zeithaml 1981), which is relevant for the hypothesized moderator effect with respect to the link between the speed of integration and market-related performance. Avoiding customer uncertainty is particularly relevant in services industries (e.g., Murray 1991). We therefore argue that the uncertainty-reducing effect of a high speed of integration is more important for firms that market services than for those that market tangible products. Thus, we put forward the following hypothesis:

H₁₁: For firms that market tangible products (as opposed to services), the effect of the (a) extent of integration on cost savings is more positive and (b) speed of integration on market-related performance after the merger or acquisition is less positive.

Table 1 presents an overview of the hypotheses related to moderating effects.

Sample and Data Collection Procedure

We used a survey methodology for data collection, which took place in 2002. The research was conducted in the German-speaking part of Central Europe (Germany, Austria, and Switzerland) with a strong focus on German firms. The initial sample consisted of horizontal M&A that took place between companies based in these countries during 1996-1999.( n4) We identified the M&A from several sources, including the Mergers & Acquisitions Database of the University of St. Gallen in Switzerland and several M&A-related European business magazines.

We initially identified 3360 reported horizontal M&A. On the basis of telephone calls with marketing/sales managers in the acquiring firm, we excluded those M&A in which the two firms marketing activities remained totally independent. In the same telephone call, we obtained the names of a senior executive or head of marketing and/or sales with responsibility for the PMI in the acquiring company.( n5) We identified managers responsible for the marketing integration of a total sample of 1483 M&A and sent questionnaires to them. We made follow-up telephone calls to verify the contact name and to encourage response.

On the basis of the follow-up telephone calls and undeliverable mail, we found 181 firms that were inappropriate for the study. However, 232 usable questionnaires were returned, for a response rate of 17.8%. We tested nonresponse bias by comparing early versus late respondents (Armstrong and Overton 1977). We also analyzed whether the firms we initially addressed and the responding firms differed in terms of industry. Both tests indicate that nonresponse bias is not a problem. We present respondent and M&A characteristics in Table 2.

Measure Development and Assessment

We followed standard psychometric scale development procedures (Gerbing and Anderson 1988). We were guided by a review of the literature (i.e., construct definitions and existing scales used in marketing and M&A research), as well as by the results from ten field interviews with practitioners. We provide a complete list of items in the Appendix.

We assessed the extent of integration using eight items related to the extent to which the systems, structures, activities, and processes in marketing were made similar. The specific items were partly based on items used by Datta (1991). The speed of integration was measured with the same eight items focusing on the shortness of the time period needed for the integration.

We measured the construct magnitude of cost savings with nine items, including the reduction in the products and services offered, brands, marketing, and sales personnel. The content of these items was largely generated through our field interviews. The measurement of market-related performance after the merger or acquisition was based on previous conceptualizations of market-related performance in the literature (e.g., Homburg and Pflesser 2000; Irving 1995). From these studies scales, we selected those items that are particularly relevant in the context of M&A (i.e., market share and customer loyalty). The conceptualization is also in line with previous research on M&A (e.g., Capron 1999; Datta 1991).

Consistent with our definition and previous M&A research (e.g., Anand and Singh 1997), financial performance after the merger or acquisition compares merging firms' profitability before and after the merger. In line with previous M&A-related research (e.g., Datta 1991; Hunt 1990), we use return on sales as our measure.( n6, n7) We measured customer orientation of integration from the managers perspective because customers are typically unaware of the driving forces behind company decisions. Its operationalization was influenced by Narver and Slater's (1990) customer orientation scale. Relatedness of the firms market positioning refers to the extent to which the merging firms offers are similar in terms of the customer needs they satisfy, such as quality or price positioning. The construct was measured with five items based on previous research (Achrol 1992). Relative size of the acquired firm was measured with a single item and in accordance with studies by Capron and Hulland (1999) and Datta (1991). Market growth before the merger or acquisition was also measured with a single item. Finally, we used our industry measure to categorize product and service firms (see Table 2). In Table 3, we show the summary statistics and correlations among constructs and moderator variables.

We analyzed measurement issues for each factor individually, and we provide the corresponding results in the Appendix. The results indicate acceptable psychometric properties in terms of internal consistency for all constructs. The average variances extracted and composite reliabilities are greater than the recommended threshold values of .5 and .6, respectively (Bagozzi and Yi 1988). In addition, all individual item reliabilities are greater than the required value of .4 (Bagozzi and Baumgartner 1994). Moreover, coefficient alpha values are greater than the threshold value of .7 (Nunnally 1978).

We assessed discriminant validity by performing pair-wise chi-square difference tests in which we compared a perfect correlation model with a model in which the two factors were allowed to correlate freely (Jöreskog and Sörbom 1982). We also applied the procedure suggested by Fornell and Larcker (1981). Both procedures indicated discriminant validity between all pairs of constructs.

Further Measure Validation Through Additional Data

Because performance assessments based on self-reported data can be problematic, we conducted two additional data collections to ensure the validity of the financial performance measure and the customer loyalty measure, a component of market-related performance.

For the validation of the financial performance measure, we identified all M&A in our sample for which both firms were publicly traded before the transaction. In this case, profitability information before the merger or acquisition is publicly available for both the acquiring and the target firms. These M&A represent 18.6% (i.e., 43 cases) of our total sample. We obtained sales volume and profitability information (return on sales) for these cases from a business information broker. Using this information, we calculated the following profitability development index (PDI).

The year in which the transaction took place is t = 0. For each year preceding the merger or acquisition (i.e., t = 3, t = 2, t = 1), we computed the relative magnitude of the two firms in terms of sales volume (AF = acquiring firm, TF = target firm). The specific formulas are as follows:

( 1) m_A,t = SAF_t/SAF_t + STF_t (t = -3, -2, -1), and

m_T,t = 1 - m_A,t,

where SAF refers to the sales volume of the acquiring firm and STF denotes the sales volume of the target firm. We then computed the weighted average profitability of these firms for each year. This weighted average profitability is

( 2) PAF_t x m_A,t + PTF_t x m_T,t

for year t, where PAF_t and PTF_t denote the profitability of the acquiring firm and the target firm, respectively, in year t. On the basis of these data, we computed the PDI as

( 3) [Multiple line equation(s) cannot be represented in ASCII text]

where PCF_t denotes the profitability of the combined firm in year t. Thus, this index compares the average profitability of the combined firm during the three years after the transaction with the weighted average profitability of both firms during the three years before the transaction. A positive value of PDI indicates an increase in profitability. For example, if the average return on sales is 5% before the merger and 7% after the merger, PDI equals 2.

We then correlated the PDI values with managers' initial evaluations of profitability after the merger or acquisition (i.e., the perceptual item used in our study). The correlation of .704 (p < .01) indicates the high validity of the managerial assessment of postmerger financial performance.

As a basis for the validation of the customer loyalty measure, we selected two industries from the total sample: banks/insurance and machinery. Two weeks after we completed the data collection, we contacted respondents in both industries by telephone. We asked the managers to name three to five customers of the combined firm that had had a business relationship with one of the firms at the time of the merger or acquisition. Forty-eight respondents named at least three customers. We then interviewed customers of 15 firms (2 customers per firm) in both industries by telephone, resulting in a total of 60 customer interviews. We asked the customers about their loyalty toward the specific firm that had undergone the merger or acquisition and how their loyalty had changed since the merger or acquisition.( n8) We then correlated the loyalty assessments by customers with managers initial evaluations of customer loyalty after PMI. We obtained high correlations (.757 for banks/insurance and .677 for machinery; p < .01 for both industries), which indicate the high validity of the managerial assessments of customer loyalty.

After establishing the structure of the measurement model, we analyzed the overall causal model (reported in Figure 2) using LISREL 8 (Jöreskog and Sörbom 1996). The fit statistics indicate an adequate fit of the model with the data (2/degrees of freedom = 3.76, goodness-of-fit index = .94, adjusted goodness-of-fit index = .93, confirmatory fit index = .95, root mean square error of approximation = .07).

Results Related to Main Effects

In Figure 2, we summarize the results of our hypotheses testing. As we show, all six hypotheses are supported by our empirical findings. More specifically, each of the parameter estimates is significant at least at the .05 level. The strongest effects are observed with respect to the link between market-related performance and financial performance and between the extent of integration and magnitude of cost savings.( n9)

The extent of integration has differential (i.e., positive and negative) indirect effects on financial performance. To explore the impact of the extent of integration on financial performance further, we computed the magnitudes of the three indirect effects and the resulting total effect. The positive effect through cost savings (.57 x .07 = .039) is confronted with two negative effects through market-related performance (.23 x .67 = .154) and a more complex causal chain through market-related performance and cost savings (.23 x .12 x .07 = .001). Thus, the total effect (.116) is negative. In other words, the positive effect of the extent of integration on financial performance (through cost savings) is reversed to a negative total effect through the negative market-related consequences of integration.

Results Related to Moderating Effects

After we found support for the main effects, we assessed the influence of the five moderator variables. We conducted separate median splits in our sample based on the values of each individual moderator. We performed multiple group LISREL (Jöreskog and Sörbom 1996) to compare the two subsamples (low versus high values of the moderator variable) for each moderator variable individually.

In the first step, we compared two models that differ with respect to all effects we consider in our moderator analysis (γ₁₁, γ₂₁, γ₂₂ for H₇, H₉, H₁₁ and γ₁₁, γ₂₂ for H₈, H₁₀). One model restricts these parameters to be equal across subsamples, whereas the more general model allows these parameters to vary across groups. Because these are nested models, the χ² value will always be lower for the general model than for the restricted model. The question is whether the improvement in χ² when moving from the restricted to the more general model, is significant. This improvement would indicate that the moderator variable has an effect on the relationships under consideration. The χ² statistic is significant for all five moderator variables (Table 4); thus, each moderator variable has some relevance in the context of our study.

In the next step, we conducted more detailed analyses and compared, for each moderator, two models that differ only with respect to one specific effect (γ₁₁, γ₂₁, or γ₂₂. In this case, one model restricts the specific parameter to be equal across subsamples, whereas the more general model allows the specific parameter to vary across groups. Again, the question is whether the improvement in χ² is significant. This improvement would indicate differential effects in the two subsamples and thus a moderator effect. We now discuss the results related to each moderator variable.

With respect to the customer orientation of integration, the χ²-difference values indicate a significant moderator effect for H7. In addition, the parameter estimates obtained in the two subsamples support our theoretical reasoning because the effect is consistently weaker for high than for low levels of customer orientation. In summary, H₇ is fully supported.

H₈, which pertains to the moderating effects of the relatedness of the merging firms market positioning, is also supported. The results show that the moderating effects are not strong enough to moderate the effects under consideration. In the conditions in which we predict and find a weaker effect, the effect is still significant.

With respect to the relative size of the acquired firm, the χ²-difference values indicate a significant moderator effect for H_9a and H_9c. Furthermore, our theoretical reasoning is supported by the parameter estimates obtained in the two subsamples; in both cases, the effect is stronger for high than for low levels of relative size. However, we fail to find support for H_9b, for which the χ²-difference is not significant. A possible explanation for this nonfinding is that our hypothesized strengthening moderator effect is compensated for by other weakening moderator effects. For example, the acquisition of a relatively large target firm may yield a large number of competent managers and staff members, and gaining this resource may weaken the negative consequences in the marketplace. Thus, H9 is only partially supported.

With regard to market growth before the merger or acquisition, H_10a is supported by our findings. The χ² difference is significant, and the effect of the extent of integration on cost savings (γ₁₁) is stronger for low than for high market growth. We fail to find support for H_10b because, contrary to our hypothesis, the negative link between the extent of integration and market-related performance is weaker in the case of high market growth. Our argument for our hypothesis was that higher switching barriers, fewer alternative suppliers, and fewer competitors attempts to promote customer switching in mature markets (i.e., markets with a low growth rate) would weaken negative customer reactions. A possible explanation for the finding is that the customer uncertainty created through a merger or acquisition is so strong that customers are highly receptive to competitors offers and that switching barriers cannot prevent customers from reacting negatively.

Our findings provide support for H_10c. We observe a negative effect for the link between the speed of integration and market-related performance in the case of low market growth. A possible explanation is that in the case of low market growth, the (positive) uncertainty-reducing effect of speed is outweighed by other mechanisms. For example, fast integration decisions involve the risk of making wrong decisions that can result in premature solutions (Schweiger and Walsh 1990, p. 61). In summary, we obtain partial support for H₁₀.

For the distinction between product and service firms, both χ²-difference values indicate a significant moderator effect. According to the parameter estimates obtained in the two subsamples, H₁₁ is fully supported. The influence of the product/service distinction is not strong enough to moderate away one of the effects.

Research Issues

Previous M&A-related research has largely neglected marketing issues despite their immense importance. Against this background, we adopted a marketing perspective toward M&A. We believe that our study advances academic knowledge in our discipline in several ways.

First, in line with the study by Capron and Hulland (1999), we contribute to bridging the gap between M&A-related research and marketing research. Our findings provide support for the idea that PMI related to marketing is highly relevant for M&A performance.

Second, we find that the extent of marketing integration has positive consequences in terms of cost savings, which are outweighed, however, by negative market-related consequences. Thus, our research provides a deeper understanding of the performance implications of the postmerger extent of integration than has most previous M&A research (e.g., Larsson and Finkelstein 1999). However, it is worth emphasizing that this finding relates to marketing integration and cannot be readily generalized to integration in other areas (e.g., manufacturing, logistics, accounting).

Third, we show that the speed of integration, a construct neglected in previous M&A research, is beneficial in terms of market-related performance. This finding extends previous conceptual discussions about the outcomes of speed in PMI (e.g., Buono and Bowditch 1989; Schweiger and Walsh 1990). Whereas we agree with these purely conceptual studies that speed can have negative outcomes, our findings show that for market-related aspects of M&A performance, speed of marketing integration is beneficial. Our reasoning--that fast integration processes can limit customer uncertainty and therefore enhance market performance--is supported by our data.

Fourth, we show that there are contingency factors that systematically strengthen or weaken the relationships under consideration. To the best of our knowledge, our study is the first to address theoretically and show empirically that the links between marketing integration and integration outcomes are not equally strong in every situation. On a more general level, we draw the conclusion that marketing research and research in the strategy field should consider moderating effects to a greater extent than they have previously. Although the number of articles analyzing moderating effects is growing, many empirical studies are still restricted to main effects. This approach is likely to miss a significant part of the real structure of the relationships among the constructs under consideration.

Fifth, we find that the negative market-related outcomes of a high level of integration are much weaker in the case of high customer orientation (H_7b). Together with the observation that many PMI phases are characterized by a strong internal orientation and a lack of customer orientation, this finding provides a possible explanation of why so many M&A fail. We suggest that this is the case because of a lack of customer orientation during PMI.

Sixth, our findings indicate that market-related performance is a much more important driver of financial performance after the merger or acquisition than are cost savings. This finding is potentially relevant for other areas of marketing research. Many marketing decisions involve a tradeoff between saving costs and building strong market positions. Our empirical study reveals a specific marketing approach (i.e., extensive marketing integration in the PMI phase) that enables firms to save costs but only at the expense of their market position, which is more important in driving financial performance than are the cost reductions. Although this finding cannot be readily generalized to other marketing approaches, we conclude that marketing researchers should be sensitive to the market position/ cost trade-offs of specific marketing approaches. This issue might be interesting to study with respect to the cross-country standardization of products and brands in the field of international marketing.

Limitations and Avenues for Further Research

Several limitations of our study must be mentioned. These limitations also provide avenues for further research. First, the sample of our study is restricted to a specific region in Europe. Additional research might use cross-country or cross-continent comparisons to study PMI.

Second, our findings reveal that positive market-related effects are more important in driving financial performance than is cost reduction, a finding based on the total sample. It may still be the case that for some mergers or acquisitions, cost savings are a more important driver of performance. Further research might address this issue in more detail.

Third, our study uses data obtained from customers to a limited extent. Additional studies in this field should use customer-based data to a greater extent than we did to achieve a deeper understanding of the processes that drive customer reactions to M&A. More specifically, research might measure the extent and speed of integration (with respect to the integration of customer-oriented activities such as product offers, brands, and sales forces) and market-related performance (with respect to customer loyalty) using customer survey data. Similarly, the moderator variable customer orientation of integration could be measured from the customers' perspective. This perspective would generate interesting and more general insights into the perception processes of customers during or after a merger or acquisition.

Fourth, our study uses perceptual measures for the magnitude of cost savings and extent of integration constructs. To make more precise financial predictions of how a certain reduction in costs affects return on sales, additional research might use objective measures of this construct. We also suggest that further research use objective measures of the extent of integration, which would require a content analytic approach to data collection in which researchers specifically consider the decisions made in the course of integration.

Fifth, our study merely considers the moderating effects of the customer orientation of integration. It is plausible that customer orientation of integration has a direct effect on cost savings and market-related performance. Thus, further work could analyze the additional effects of a customer orientation in PMI. In addition, it would be worthwhile to study how the customer orientation, in general, of either firm before the transaction affects the integration outcomes.

Sixth, our study considers the extent and speed of integration as independent variables. Additional studies could analyze which variables drive the extent and speed of marketing integration. As an example, researchers could analyze how the relative size of the acquired firm or relatedness of the firms market positioning affects the extent and speed of integration.

Finally, our study focuses on cost savings and market-related performance as drivers of financial performance. An interesting issue for further research would be to study additional variables that influence postmerger performance, such as how increased prices as a consequence of a merger or acquisition affect financial performance.

Managerial Implications

A first managerial implication results from the finding that marketing integration affects financial performance negatively because the negative effects of market-related performance outweigh the positive effects of cost savings. Thus, managers should be cautious about postmerger marketing integration that is driven by the goal of reducing costs. Although it is realistic to expect cost savings from a high extent of integration, our findings show that there is a price to be paid for these cost savings on the market side. Our study also shows managers how to weaken these negative market-related consequences through a high level of customer orientation of integration, because this variable is a significant moderator of this link (H_7b). Thus, the specific implication for managers is that they should devote a lot of attention to customer-related issues when making decisions about PMI. For example, integration activities should be strongly guided by the needs of the merged firm's customers with respect to the product offer and customer service. Systematic customer surveys can be helpful in this context.

A second managerial implication is that, in general, managers should strive for speed in integrating their marketing operations after a merger or acquisition. Our research provides managers with information about the circumstances in which speed is particularly beneficial. For example, the effect of integration speed on market-related performance is particularly strong when the relatedness of the merging firms market positioning is low, the relative size of the acquired firm is high, and the firms are service firms.

To give managers a more concrete picture with respect to the speed of integration, in Table 5, we provide detailed information about integration speed with respect to all the issues covered in the construct. These objective numbers can be used as benchmarking information by managers.

Finally, the analysis of market-related performance and cost savings as drivers of financial M&A performance leads to straightforward managerial implications. Managers involved in PMI should strongly focus on market-related issues. For example, concrete managerial actions during the integration phase could include the implementation of customer satisfaction--based incentive systems or continuous customer feedback instruments. A predominant cost focus during PMI is likely to leave important performance potentials unexploited and lead to risks that produce negative outcomes in the market that cannibalize the success on the cost side.

Our research provides strong evidence for the importance of studying M&A from a marketing perspective. We show that the marketing integration process is highly relevant in driving ultimate M&A performance. We also show that market-related performance is far more important than are cost savings. Against this background, we conclude that the marketing discipline should devote more research attention to marketing issues in the context of M&A. It has been argued that the marketing discipline has had a limited impact on the field of strategic management (e.g., Day 1992). The discipline s influence can be increased only if key issues in corporate strategy are addressed from a marketing perspective by marketing scholars. We propose that studying drivers of M&A performance is one of these key issues that must not be neglected by marketing researchers. It is our hope that our study, following the pioneering work by Capron and Hulland (1999), contributes to the emergence of a research stream on M&A in marketing.

The authors thank the Stiftung Rheinische Hypothekenbank for supporting the research reported in this article and the University of St. Gallen for providing access to its mergers and acquisitions database.

( n1) The term "integration" has been used with different meanings in different literature streams. As an example, literature in the field of international marketing (e.g., Birkinshaw, Morrison, and Hulland 1995; Zou and Cavusgil 2002) has used integration specifically in terms of the integration of marketing activities across countries. Our use of the term was guided by previous research in the M&A field and is consistent with this work (Pablo 1994; Shrivastava 1986).

( n2) H₁ is not tautological in nature because there is a clear conceptual distinction between the independent construct and the dependent construct. A high level of integration is not identical to a high level of cost savings for three reasons. First, a high level of integration does not automatically imply that something positive has been achieved. Rather, a high level of integration merely means that a high level of similarity has been achieved (for better or worse) between the two organizations with respect to a certain marketing aspect. For example, if two information systems are replaced by one information system (i.e., a high level of integration is achieved), the chosen integrated information system still may be complex to handle and therefore may increase costs. Second, a high level of integration in some areas creates the potential for cost reductions. However, this potential may or may not be exploited by managers. Third, a high level of integration may be driven by the aim to increase effectiveness rather than reduce costs. For example, if one company has better procedures in new product development or pricing, integration would primarily result in increases in effectiveness.

( n3) We do not find any compelling argument for either a positive or a negative direct impact of the extent of integration on financial performance after the merger or acquisition. Therefore, we do not put forward a hypothesis corresponding to this possible link. However, on the basis of our empirical results, we explore the magnitude of the indirect effects between extent of integration and financial performance through the magnitude of cost savings and market-related performance.

( n4) We chose this period to exclude recent M&A in which the integration process was still in an early stage and had not yet led to any significant outcomes at the time of the survey, as well as older transactions for which it is difficult to obtain detailed information about integration activities because of managerial turnover. We restricted our search to horizontal M&A (M&A that take place in the same industry, often between direct competitors) because the issue of marketing integration essentially only comes up in horizontal M&A.

( n5) We chose managers from the acquiring firm because they tend to be the most knowledgeable about the PMI. In addition, because of managerial turnover, it is often impossible to identify former executives from the target firm (Walsh 1988).

( n6) We used perceptual measures for market share and return on sales because respondents are typically reluctant to provide or do not know the precise numbers on performance constructs (Dess and Robinson 1984). Both problems may lead to respondents breaking up the response process, which reduces the achieved response rate (for similar arguments, see Datta 1991). Note, however, that we validated the perceptual profitability measure with objective data from an independent source.

( n7) Ultimate M&A success also depends on the price that was paid for the acquiree. We did not include this aspect in our framework because managers are typically reluctant to provide this information. In addition, for the majority of transactions, price information is not publicly available.

( n8) We assessed customer loyalty in the validation sample using multiple items, such as the intentions to purchase from the specific firm again, expand the relationship with the firm, and recommend the firm to others. The specific items were partly based on items used by Cannon and Homburg (2001) and Zeithaml, Berry, and Parasuraman (1996).

( n9) On the basis of a suggestion by an anonymous reviewer, we also analyzed an alternative model in which the relationship under consideration runs in the opposite direction (i.e., magnitude of cost savings is hypothesized to affect market-related performance). With respect to this alternative model, our findings indicate that the effect of the magnitude of cost savings on market-related performance is not significant (β₂₁ = .04, p > .1). In addition, the overall fit of the model is poorer than the fit of the model we propose.

Hypotheses Related to Moderating Effects

Legend for Chart:

A - Moderator Variables
B - Main Effects Subject to Moderation Extent of integration
    → cost savings Expected Direction of Main Effects
    Positive
C - Main Effects Subject to Moderation Extent of integration
    → market-related performance Expected Direction of Main
    Effects Negative
D - Main Effects Subject to Moderation Speed of integration
    → market-related performance Expected Direction of Main
    Effects Positive

      A                                        B

                                               C

                                               D

Customer orientation of          Effect is less positive for
integration (COI)                high values of COI (H7a)

                                 Effect is less negative for
                                 high values of COI (H7b)

                                 Effect is less positive for
                                 high values of COI (H7c)

Relatedness of the firms'        Effect is more positive for
market positioning (RMP)         high values of RMP (H8a)

                                 --

                                 Effect is less positive for
                                 high values of RMP (H8b)

Relative size of the acquired    Effect is more positive for
firm (RS)                        high values of RS (H9a)

                                 Effect is more negative for
                                 high values of RS (H9b)

                                 Effect is more positive for
                                 high values of RS (H9c)

Market growth before the         Effect is less positive for
merger or acquisition(MG)        high values of MG (H10a)

                                 Effect is more negative for
                                 high values of MG (H10b)

                                 Effect is more positive for
                                 high values of MG (H10c)

Product (PF) versus service      Effect is more positive for PF
firms (SF)                       (H11a)

                                 --

                                 Effect is less positive for PF
                                 (H11b)

Sample Comparison (232 Cases)

Legend for Chart:

B - %

        A                                                   B

A: Industry

Banks and insurance                                         38
Machinery                                                   23
Food and packaged goods                                     11
Chemicals                                                    8
Printing and publishing services                             6
Automotive components                                        5
Pharmaceuticals                                              3
Other                                                        4
Missing                                                      2

B: Annual Revenues
(of the consolidated business)

<$25 million                                                23
$25-$49 million                                             13
$50-$99 million                                             15
$100-$249 million                                            7
$250-$499 million                                           11
$500-$1,000 million                                         14
>$1,000 million                                             16
Missing                                                      1

C: Position of Respondents

Managing director, chief executive officer, vice
 president of region, head of strategic business unit       67
Vice president marketing, vice president sales, vice
 president marketing and sales                              18
Sales manager, product manager                               6
Head of M&A                                              5
Other                                                        2
Missing                                                      2

D: Relative Size of Target to Acquirer
(annual revenues)

<25%                                                        38
25%-49%                                                     23
50%-74%                                                     13
75%-100%                                                    10
>100%                                                       14
Missing                                                      2

Correlations and Summary Statistics

Legend for Chart:

A - Variables
B - Correlations 1
C - Correlations 2
D - Correlations 3
E - Correlations 4
F - Correlations 5
G - Correlations 6
H - Correlations 7
I - Correlations 8
J - Correlations 9

     A
                     B        C        D        E        F
                              G        H        I        J

1. Extent of integration
                   1.00

2. Speed of integration
                    .273    1.00

3. Magnitude of cost savings
                    .297     .209    1.00

4. Market-related performance after the merger or acquisition
                   -.075     .105     .121    1.00

5. Financial performance after the merger or acquisition
                   -.105     .226     .079     .520    1.00

6. Customer orientation of integration
                    .357     .304     .140     .256     .184
                            1.00

7. Relatedness of the firms' market positioning
                    .484     .261     .139    -.071    -.044
                             .236    1.00

8. Relative size of the acquired firm
                   -.132     .007     .025    -.087     .011
                            -.056    -.250    1.00

9. Market growth before the merger or acquisition
                    .128     .129    -.060     .096    -.054
                             .129     .094    -.103    1.00

Summary Statistics

Range
                     1-7      1-5      1-7      1-7      1-7
                              1-7      1-7      1-5      1-9

Number of items
                       8        8        9        2        1
                                3        5        1        1

Mean
                    4.19     3.79     2.62     4.04     4.75
                             4.82     4.77     3.62     5.17

Standard deviation
                    1.72      .98     1.25      .96     1.30
                             1.72     1.30     1.44     1.21

Results of Multiple Group Analysis

Legend for Chart:

A - Moderator Variable
B - Main Effect
C - Hypothesized Moderator Effect
D - Low Value of Moderator
E - High Value of Moderator
F - χ² Difference (ΔDF = 1)

   A
           B

               C

               D                         E

Customer
orientation of
integration
(COI)
           Extent of integration → cost savings
           (positive)

               H7a: Effect is less positive for high values
               of COI

               γ11 = .64     γ11 = .47
               (t = 13.10)               (t = 6.50)

               Δχ² = 27.83(**)

           Extent of integration → market-related
           performance (negative)

               H7b: Effect is less negative for high values
               of COI

               γ21 = -.55     γ21 = .10
               (t = -8.42)               (t = 1.61)

               Δχ² = 4.83(*)

           Speed of integration → market-related
           performance (positive)

               H7c: Effect is less positive for high values
               of COI

               γ22 = .18      γ22 = -.08
               (t = 4.03)                (t = -1.28)

               Δχ² = 9.37(**)

Δχ² for all gammas set equal across
subgroups (ΔDF = 3): 43.79(**)

Relatedness of
the firms'
market
positioning
(RMP)
           Extent of integration → cost savings
           (positive)

               H8a: Effect is more positive for high values
               of RMP

               γ11 = .38     γ11 = .69
               (t = 8.79)                (t = 13.32)

               Δχ² = 16.83(**)

           Speed of integration → market-related
           performance (positive)

               H8b: Effect is less positive for high values
               of RMP

               γ22 = .24     γ22 = .18
               (t = 5.26)                (t = 3.02)

               Δχ² = 3.80(*)

Δχ² for all gammas set equal across
subgroups (ΔDF = 2): 20.63(**)

Relative size of
the acquired
firm (RS)
           Extent of integration → cost savings
           (positive)

               H9a: Effect is more positive for high values
               of RS

               γ11 = .15     γ11 = .78
               (t = 4.71)                (t = 13.18)

               Δχ² = 46.41(**)

           Extent of integration → market-related
           performance (negative)

               H9b: Effect is more negative for high values
               of RS

               γ21 = -.33    γ21 = -.14
               (t = -7.31)               (t = -1.32)

               Δχ² = 1.80

           Speed of integration → market-related
           performance (positive)

               H9c: Effect is more positive for high values
               of RS

               γ22 = .08     γ22 = .60
               (t = 2.41)                (t = 4.73)

               Δχ² = 9.49(**)

Δχ² for all gammas set equal across
subgroups (ΔDF = 3): 46.92(**)

Market growth
before the
merger or
acquisition(MG)
           Extent of integration → cost savings
           (positive)

               H10a: Effect is less positive for high
               values of MG

               γ11 = .71     γ11 = .31
               (t = 14.91)               (t = 10.14)

               Δχ² = 25.49(**)

           Extent of integration → market-related
           performance (negative)

               H10b: Effect is more negative for high
               values of MG

               γ21 = -.25    γ21 = -.17
               (t = -4.82)               (t = -4.89)

               Δχ² = 5.85(**)

           Speed of integration → market-related
           performance (positive)

               H10c: Effect is more positive for high
               values of MG

               γ22 = -.13    γ22 = .42
               (t = -3.77)               (t = 6.69)

               Δχ² = 31.04(**)

Δχ² for all gammas set equal across
subgroups (ΔDF = 3): 57.67(**)

Product(PF)
versus service
firms(SF)
           Extent of integration → cost savings
           (positive)

               H11a: Effect is more positive for PF

               γ11 = .63     γ11 = .43
               (t = 13.65)               (t = 12.47)

               Δχ² = 9.13(**)

           Speed of integration → market-related
           performance (positive)

               H11b: Effect is less positive for PF

               γ22 = .19     γ22 = .35
               (t = 4.31)                (t = 6.24)

               Δχ² = 9.26(**)

Δχ² for all gammas set equal
across subgroups (ΔDF = 2): 16.23(**)

(*) p < .05.
(**) p < .01.

Detailed Frequency Distributions for Speed of Integration

Legend for Chart:

A - Items
B - Duration of Integration <6 months
C - Duration of Integration 6-12 months
D - Duration of Integration 13-18 months
E - Duration of Integration 19-24 months
F - Duration of Integration >24 months

         A                               B        C        D
                                                  E        F

1. Products/services offered           32.0%    34.5%    14.3%
                                                10.8%     8.4%

2. New product development             29.9%    25.9%    22.4%
                                                10.9%    10.9%

3. Prices                              54.2%    26.8%    10.5%
                                                 4.7%     3.8%

4. Communication                       56.4%    22.6%    10.8%
                                                 5.1%     5.1%

5. Sales system                        36.7%    27.1%    20.3%
                                                 8.7%     7.2%

6. Sales force management              38.9%    23.8%    16.1%
                                                12.4%     8.8%

7. Information systems                 37.5%    26.0%    21.2%
                                                 8.2%     7.1%

8. Internal marketing/sales support    40.8%    25.1%    16.8%
                                                 9.4%     7.9%

DIAGRAM: FIGURE 1; Framework and Constructs

DIAGRAM: FIGURE 2; Results of the Hypotheses Testing (Main Effects)

Achrol, Ravi S. (1992), The Dimensions of Marketing Channel Environments, in Advances in Distribution Channel Research,, Gary L. Frazier, ed. Greenwich, CT: Elsevier, 1-43.

Anand, Jaideep and Harbir Singh (1997), Asset Redeployment, Acquisitions and Corporate Strategy in Declining Industries, Strategic Management Journal 18 (Summer Special Issue), 99-118.

Anderson, Eugene W., Claes Fornell, and Roland T. Rust (1997), Customer Satisfaction, Productivity, and Profitability: Differences Between Goods and Services, Marketing Science 16 (2), 129-47.

Armstrong, S. and T. Overton (1977), Estimating Nonresponse Bias in Mail Surveys, Journal of Marketing Research 14 (August), 396-402.

Arndt, Johan (1979), Toward a Concept of Domesticated Markets, Journal of Marketing 43 (July), 69-75.

Bagozzi, Richard P. and Hans Baumgartner (1994), The Evaluation of Structural Equation Models and Hypothesis Testing, in Principles of Marketing Research R.P. Bagozzi, ed. Cambridge: Blackwell Publishers, 386-422.

----- and Youjae Yi ((1988), On the Evaluation of Structural Equation Models, Journal of the Academy of Marketing Science 16 (Spring), 74-94.

Barney, Jay B. (1991), Firm Resources and Sustained Competitive Advantage, Journal of Management 17 (1), 99-120.

----- (1986), "Strategic Factor Markets: Expectations, Luck, and Business Strategy," Management Science 42 (10), 1231-41.

Becker, Sally A. and Darryl Flamer (1997), "Checking out Customers' Needs After the Deal," Mergers and Acquisitions 32 (3), 8-10.

Bekier, Matthias M. and Michael J. Shelton (2002), "Keeping Your Sales Force after the Merger," The McKinsey Quarterly 4, 106 115.

Birkinshaw, Julian, Henrik Bresman, and Lars Hakanson (2000), "Managing the Post-Acquisition Integration Process: How the Human Integration and Task Integration Processes Interact to Foster Value Creation," Journal of Management Studies 37 (3), 395 425.

-----, Allen Morrison, and John Hulland (1995), "Structural and Competitive Determinants of a Global Integration Strategy," Strategic Management Journal 16 (8), 637-55.

Blau, Peter M. and Richard A. Schoenherr (1971), The Structure of Organizations New York: Basic Books.

Bowman, Douglas and Hubert Gatignon (1995), "Determinants of Competitor Response Time to a New Product Introduction," Journal of Marketing Research 32 (February), 42-53.

Bragado, Julio F. (1992), Setting the Correct Speed for PostMerger Integration, M&A Europe 5 (March/April), 24-31.

Budros, Art (1999), "A Conceptual Framework for Analyzing Why Organizations Downsize," Organization Science 10 (1), 69-82.

Buono, Anthony F. and James L. Bowditch (1989), The Human Side of Mergers and Acquisitions San Francisco: Jossey-Bass.

Buzzell, Robert D. (1981), "Are There Natural Market Structures?" Journal of Marketing 45 (October), 42 51.

Cannon, Joseph P. and Christian Homburg (2001), "Buyer Supplier Relationships and Customer Firm Costs, Journal of Marketing 65 (January), 29-43.

Capron, Laurence (1999), The Long-Term Performance of Horizontal Acquisitions, "Strategic Management Journal 20 (11), 987-1018.

-----, Pierre Dussauge, and Will Mitchell (1998), "Resource Redeployment Following Horizontal Acquisitions in Europe and North America, 1988-1992," Strategic Management Journal 19 (7), 631-61.

----- and John Hulland ((1999), "Redeployment of Brands, Sales Forces, and General Marketing Management Expertise Following Horizontal Acquisitions: A Resource-Based View," Journal of Marketing 63 (April), 41-54.

Chakrabarti, Alok K. (1990), "Organizational Factors in PostAcquisition Performance," IEEE Transactions and Engineering Management 37 (4), 259-68.

Chatterjee, Sayan (1986), "Types of Synergy and Economic Value: The Impact of Acquisitions on Merging and Rival Firms," Strategic Management Journal 7 (2), 119-39.

Clemente, Mark N. and David S. Greenspan (1997), "Keeping Customers Satisfied While the Deal Proceeds," Mergers and Acquisitions 32 (1), 24-28.

Colgate, Mark and Bodo Lang (2001), "Switching Barriers in Consumer Markets: An Investigation of the Financial Services Industry," Journal of Consumer Marketing 18 (4), 332-47.

Datta, Deepak K. (1991), "Organizational Fit and Acquisition Performance: Effects of Post-Acquisition Integration," Strategic Management Journal 12 (4), 281-97.

Davis, Rachel and L.G. Thomas (1993), "Direct Estimation of Synergy: A New Approach to the Diversity Performance Debate," Management Science 39 (11), 1334-46.

Day, George S. (1981), "The Product Life Cycle: Analysis and Applications Issues," Journal of Marketing 45 (July), 60-67.

----- (1992), "Marketing's Contribution to the Strategy Dialogue, Journal of the Academy of Marketing Science 20 (3), 323-29.

Dess, Gregory G. and Richard B. Robinson (1984), "Measuring Organizational Performance in the Absence of Objective Measures: The Case of the Privately-Held Firm and Conglomerate Business Unit," Strategic Management Journal 5 (3), 265-73.

Fornell, Claes (1992), A National Customer Satisfaction Barometer: The Swedish Experience, Journal of Marketing 56 (January), 6-21.

----- and David F.. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error," Journal of Marketing Research 18 (February), 39-50.

Gerbing, David W. and James C. Anderson (1988), "An Updated Paradigm for Scale Development Incorporating Unidimensionality and Its Assessment," Journal of Marketing Research 25 (May), 186-92.

Hagedorn, John and Geert Duysters (2002), "The Effect of Mergers and Acquisitions on the Technological Performance of Companies in a High-Tech Environment," Technology Analysis & Strategic Management 14 (1), 67-85.

Haspeslagh, Philippe C. and David B. Jemison (1991), Managing Acquisitions: Creating Value Through Corporate Renewal New York: The Free Press.

Hitt, Michael A., Robert E. Hoskisson, and Duane R. Ireland (1990), "Mergers and Acquisitions and Managerial Commitment to Innovation in M-Form Firms," Strategic Management Journal 11 (4), 29-48.

Homburg, Christian and Christian Pflesser (2000), "A Multiple-Layer Model of Market-Oriented Organizational Culture: Measurement Issues and Performance Outcomes," Journal of Marketing Research 37 (November), 449 62.

Hunt, John W. (1990), "Changing Pattern of Acquisition Behavior in Takeovers and Consequences for Acquisitions Process," Strategic Management Journal 11 (1), 69-77.

Irving, Ed (1995), "Marketing Quality Practices, doctoral dissertation, Business School, University of North Carolina Chapel Hill.

Jöreskog, Karl and Dag Sörbom (1982), Recent Developments in Structural Equation Modeling, Journal of Marketing Research 19 (November), 404-416.

----- and ----- (1996), LISREL 8: User's Reference Guide Chicago: Lawrence Erlbaum Associates.

Kalwani, Manohar U. and Narakesari Narayandas (1995), "Long-Term Manufacturer Supplier Relationships: Do They Pay Off for Supplier Firms?" Journal of Marketing 59 (January), 16-27.

Karim, Samina and Will Mitchell (2000), "Path-Dependent and Path-Breaking Change: Reconfiguring Business Resources Following Acquisitions in the U.S. Medical Sector, 1978-1995," Strategic Management Journal 21 (10/11), 1061-81.

Krishnan, Hema A. and Daewoo Park (2002), "The Impact of Work Force Reduction on Subsequent Performance in Major Mergers and Acquisitions--An Exploratory Study," Journal of Business Research 55 (4), 285-92.

Larsson, Rikard (1989), Organizational Integration of Mergers and Acquisitions Lund: Charwell-Bratt.

----- and Sydney Finkelstein ((1999), "Integrating Strategic, Organizational and Human Resource Perspectives on Mergers and Acquisitions: A Case Survey of Synergy Realization," Organization Science 10 (1), 1-26.

Marks, Mitchell L. and Philip H. Mirvis (2001), "Making Mergers and Acquisitions Work: Strategic and Psychological Preparation," Academy of Management Executive 15 (2), 80-92.

Mitchell, D. (1989), "The Importance of Speed in Post-Merger Reorganization," M&A Europe 1 (3), 44-48.

Morall, Katherine (1996), "Managing a Merger Without Losing Customers," Bank Marketing 28 (3), 18-23.

Murray, Keith B. (1991), "A Test of Services Marketing Theory: Consumer Information Acquisition Activities," Journal of Marketing 55 (January), 10-25.

Narver, John C. and Stanley F. Slater (1990), "The Effect of a Market Orientation on Business Profitability," Journal of Marketing 54 (October), 20-35.

Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change Cambridge: Belknap Press.

Nunnally, Jum C. (1978), Psychometric Theory 2d ed. New York: McGraw-Hill.

Pablo, Amy L. (1994), "Determinants of Acquisition Integration Level: A Decision-Making Perspective," Academy of Management Journal 37 (4), 803-836.

Peteraf, Margaret (1993), "The Cornerstones of Competitive Advantage: A Resource-Based View," Strategic Management Journal 14 (3), 171-91.

Phillips Lynn W., Dae R. Chang, and Robert D. Buzzell (1983), "Product Quality, Cost Position and Business Performance: A Test of Some Key Hypotheses," Journal of Marketing 47 (April), 26-44.

Pugh, Derek S., David J. Hickson, C. Robert Hinings, and C. Turner (1969), "The Context of Organization Structures," Administrative Science Quarterly 14 (1), 91 114.

Reichheld, Frederick F. and Brad Henske (1991), "The Only Sure Method of Recouping Merger Premiums," Journal of Retail Banking 8 (2), 9-17.

Schweiger, David M. and Philip K. Goulet (2000), "Integrating Mergers and Acquisitions: An International Research Review," in Advances in Mergers and Acquisitions C.L. Cooper and A. Gregory, eds. Amsterdam: JAI Press, 61-91.

----- and James P.. Walsh (1990), "Mergers and Acquisitions: An Interdisciplinary View," Research in Personnel and Human Resources Management 8, 41-107.

Seth, Anju (1990a), "Value Creation in Acquisitions: A ReExamination of Performance Issues," Strategic Management Journal 11 (2), 99-115.

----- (1990b), "Sources of Value Creation in Acquisitions: An Empirical Investigation," Strategic Management Journal 11 (6), 431-46.

Shrivastava, Paul (1986), "Post-Merger Integration," Journal of Business Strategy 7 (1), 65-76.

Singh, Harbir and Cynthia A. Montgomery (1987), "Corporate Acquisition Strategies and Economic Performance," Strategic Management Journal 8 (4), 377-86.

Singh, Jagdip, Jerry R. Goolsby, and Gary K. Rhoads (1994), "Behavioral and Psychological Consequences of Boundary-Spanning Burnout for Customer Service Representatives," Journal of Marketing Research 31 (4), 558-69.

Szymanski, David M., Dundar G. Bharadwaj, and P. Rajan Varadarajan (1993), "An Analysis of the Market Share Profitability Relationship," Journal of Marketing 57 (July), 1-18.

Tetenbaum, Toby J. (1999), "Beating the Odds of Merger & Acquisition Failure: Seven Key Practices that Improve the Chance for Expected Integration and Synergies," Organizational Dynamics 28 (2), 22-36.

Urban, David J. and Michael D. Pratt (2000), "Perceptions of Banking Services in the Wake of Bank Merger: An Empirical Study," Journal of Services Marketing, 14 (2/3), 118-31.

Walsh, James P. (1988), "Top Management Turnover Following Mergers and Acquisitions," Strategic Management Journal 9 (2), 173-83.

Wasson, Chester R. (1978), Dynamic Competitive Strategy and Product Life Cycles. St. Charles, MO: Challenge Books.

Wernerfelt, Birger (1984), "A Resource-Based View of the Firm," Strategic Management Journal 5 (2), 4-12.

Zeithaml, Valarie A. (1981), "How Consumer Evaluation Processes Differ Between Goods and Services," in Marketing of Services James H. Donnelly and William George, eds. Chicago: American Marketing Association, 186 90.

-----, Leonard L. Berry, and A. Parasuraman (1996), "The Behavioral Consequences of Service Quality," Journal of Marketing 60 (April), 31-46.

-----, A. Parasuraman, and Leonard L. Berry (1985), "Problems and Strategies in Services Marketing," Journal of Marketing 49 (January), 33-46.

Zou, Shaoming and S. Tamer Cavusgil (2002), "The GMS: A Broad Conceptualization of Global Marketing Strategy and Its Effect on Firm Performance," Journal of Marketing 66 (October), 40-56.

Scale Items for Construct Measure

Legend for Chart:

A - Construct
B - Items
C - Individual Item Reliability(a)
D - Average Variance Extracted(a)
E - Composite Reliability/Coefficient Alpha(b)

     A

         B

                C       D         E

Extent of Integration
(seven-point rating
scale with anchors
1 = no integration,
4 = partial
integration, and 7 =
complete
integration)

         To what extent were the following aspects made
         similar between the two firms after the merger
         or acquisition?

                       .74     .96/.83

         • Products/services offered (e.g., harmonization
         of product ranges, brand names)

               .72

         • New product development

               .40

         • Prices (e.g., harmonization of the
         price positioning)

               .82

         • Communication (e.g., harmonization
         of advertisement)

               .74

         • Sales system (e.g., harmonization of sales
         channels, sales partners, sales offices)

               .83

         • Sales force management (e.g., harmonization
         of the incentive and provision systems)

               .85

         • Information systems (e.g., harmonization of the
         marketing/sales information systems)

               .80

         • Internal marketing/sales support

               .80

Speed of Integration
(five-point rating
scale with 1 = more
than 24 months, 2 =
19-24 months, 3 =
13-18 months, 4 =
6-12 months, and
5 = less than 6
months)

         How long did it take to complete the intended
         integration of the following aspects?

                       .58     .89/.82

         • Products/services offered (e.g.,
         harmonization of product ranges, brand names)

               .53

         • New product development

               .40

         • Prices (e.g., harmonization of the
         price positioning)

               .58

         • Communication (e.g., harmonization
         of advertisement)

               .59

         • Sales system (e.g., harmonization of
         sales channels, sales partners, sales offices)

               .52

         • Sales force management (e.g., harmonization
         of the incentive and provision systems)

               .66

         • Information systems (e.g., harmonization
         of the marketing/sales information systems)

               .62

         • Internal marketing/sales support

               .76

Magnitude of Cost
Savings
(seven-point rating
scale with anchors
1= not reduced at
all and 7 = strongly
reduced)

         To what extent were the following resources
         reduced as a result of the merger or acquisition?
         (Please refer to both firms together.)

                       .50     .89/.85

         • Products offered

               .43

         • Services offered

               .41

         • Brands

               .49

         • Strategic business units

               .54

         • Sales channels

               .47

         • Production locations

               .45

         • Sales offices

               .65

         • Total employees in marketing

               .66

         • Total employees in sales

               .65

Market-Related
Performance After
the Merger or
Acquisition
(seven-point rating
scale with anchors
1= significant
decline, 4 = more or
less the same, and
7 = significant
increase)

         Compared to the situation before the merger or
         acquisition, please indicate how have the merging
         firms together performed with respect to market
         share (please consider the sum of both companies).

                --      --     -- /.76

(seven-point rating
scale with anchors
1= clearly worse,
4= competition
level, and 7 = clearly
better)

         Relative to your competitors, how have the merging
         firms together performed with respect to retaining
         existing customers after the integration was completed

                --

Financial
Performance After
the Merger or
Acquisition
(seven-point rating
scale with anchors
1= significant
decline, 4 = more or
less the same, and
7 = significant
increase)

         Compared to the situation before the merger or
         acquisition, please indicate how the return on
         sales has developed (please consider the sum of
         both companies).

                --      --          --

Customer
Orientation of
Integration
(seven-point rating
scale with anchors
1= strongly
disagree and 7 =
strongly agree)

         Please indicate the degree to which you agree with
         the following statements concerning the fundamental
         orientation of the integration process after the merger
         or acquisition.

                       .85     .95/.93

         • Customer needs were decisive for the realization
         of integration activities.

               .80

         • Customer needs were the central focus during the
         integration process.

               .99

         • Integration activities were focused on the
         increase of customer benefit.

               .77

Relatedness of the
Firms' Market
Positioning
(seven-point rating
scale with anchors
1 = strongly
disagree and 7 =
strongly agree)

         Please indicate the degree to which you agree with
         the following statements concerning the market
         positioning of your company and the other company
         before the M/A?

                       .53     .85/.81

         • The products/services of both companies were
         focused on the satisfaction of the same customer
         needs.

               .49

         • The products/services of both companies
         were mainly based on the same technologies.

               .59

         • The products/services of both companies
         were mainly identical in quality.

               .51

         • The products/services of both companies
         were mainly sold through the same distribution
         channels.

               .55

         • The price positioning of the offers of
         both companies was mainly identical.

               .50

Relative Size of the
Acquired Firm
(five-point-rating
scale with 1 =
<25%, 2 =
25%-49%, 3 =
50%-74%, 4 =
75%-100%, and 5 =
>100%)

         Please indicate the relative proportion of the
         acquired firm's annual sales in comparison to
         your firm's sales in the year before the merger
         or acquisition (in the line of business concerned).

                --      --          --

Market Growth
Before the Merger or
Acquisition
(nine-point-rating
scale with 1 =
decrease >15%, 2 =
decrease 5%-15%,
3 = decrease up to
5%, 4 = increase up
to 5%, 5 = increase
5%-10%, 6 =
relatively constant,
7= increase
10%-20%, 8 =
increase 20%-30%,
and 9 = increase
>30%)

         Please indicate the average annual growth rate
         (over the last three years before the merger or
         acquisition) of the total demand in the market
         concerned.

                --      --          --

(a) Individual item reliability and average variance extracted
are reported when there are more than two items.

(b) Reports coefficient alpha (if more than one item) and
composite reliability (if more than two items).

~~~~~~~~

By Christian Homburg and Matthias Bucerius

Christian Homburg is Professor of Business Administration and Marketing and Chair of the Marketing Department, University of Mannheim

Matthias Bucerius was a doctoral student at this department. Currently, he is a management consultant in the chemical and pharmaceutical industry (e-mail: matthias@bucerius.de).

A Strategic Framework for Customer Relationship Management 

In this article, the authors develop a conceptual framework for customer relationship management (CRM) that helps broaden the understanding of CRM and its role in enhancing customer value and, as a result, shareholder value. The authors explore definitional aspects of CRM, and they identify three alternative perspectives of CRM. The authors emphasize the need for a cross-functional, process-oriented approach that positions CRM at a strategic level. They identify five key cross-functional CRM processes: a strategy development process, a value creation process, a multichannel integration process, an information management process, and a performance assessment process. They develop a new conceptual framework based on these processes and explore the role and function of each element in the framework. The synthesis of the diverse concepts within the literature on CRM and relationship marketing into a single, process-based framework should provide deeper insight into achieving success with CRM strategy and implementation.

Over the past decade, there has been an explosion of interest in customer relationship management (CRM) by both academics and executives. However, despite an increasing amount of published material, most of which is practitioner oriented, there remains a lack of agreement about what CRM is and how CRM strategy should be developed. The purpose of this article is to develop a process-oriented conceptual framework that positions CRM at a strategic level by identifying the key cross-functional processes involved in the development of CRM strategy. More specifically, the aims of this article are

• To identify alternative perspectives of CRM,

• To emphasize the importance of a strategic approach to CRM within a holistic organizational context,

• To propose five key generic cross-functional processes that organizations can use to develop and deliver an effective CRM strategy, and

• To develop a process-based conceptual framework for CRM strategy development and to review the role and components of each process.

We organize this article in three main parts. First, we explore the role of CRM and identify three alternative perspectives of CRM. Second, we consider the need for a cross-functional process-based approach to CRM. We develop criteria for process selection and identify five key CRM processes. Third, we propose a strategic conceptual framework that is constructed of these five processes and examine the components of each process.

The development of this framework is a response to a challenge by Reinartz, Krafft, and Hoyer (2004), who criticize the severe lack of CRM research that takes a broader, more strategic focus. The article does not explore people issues related to CRM implementation. Customer relationship management can fail when a limited number of employees are committed to the initiative; thus, employee engagement and change management are essential issues in CRM implementation. In our discussion, we emphasize such implementation and people issues as a priority area for further research.

The term "customer relationship management" emerged in the information technology (IT) vendor community and practitioner community in the mid-1990s. It is often used to describe technology-based customer solutions, such as sales force automation (SFA). In the academic community, the terms "relationship marketing" and CRM are often used interchangeably (Parvatiyar and Sheth 2001). However, CRM is more commonly used in the context of technology solutions and has been described as "information-enabled relationship marketing" (Ryals and Payne 2001, p. 3). Zablah, Beuenger, and Johnston (2003, p. 116) suggest that CRM is "a philosophically-related offspring to relationship marketing which is for the most part neglected in the literature," and they conclude that "further exploration of CRM and its related phenomena is not only warranted but also desperately needed."

A significant problem that many organizations deciding to adopt CRM face stems from the great deal of confusion about what constitutes CRM. In interviews with executives, which formed part of our research process (we describe this process subsequently), we found a wide range of views about what CRM means. To some, it meant direct mail, a loyalty card scheme, or a database, whereas others envisioned it as a help desk or a call center. Some said that it was about populating a data warehouse or undertaking data mining; others considered CRM an e-commerce solution, such as the use of a personalization engine on the Internet or a relational database for SFA. This lack of a widely accepted and appropriate definition of CRM can contribute to the failure of a CRM project when an organization views CRM from a limited technology perspective or undertakes CRM on a fragmented basis.

The definitions and descriptions of CRM that different authors and authorities use vary considerably, signifying a variety of CRM viewpoints. To identify alternative perspectives of CRM, we considered definitions and descriptions of CRM from a range of sources, which we summarize in the Appendix. We excluded other, similar definitions from this list.

An important aspect of the CRM definition that we wanted to examine was its association with technology. This is important because CRM technology is often incorrectly equated with CRM (Reinartz, Krafft, and Hoyer 2004), and a key reason for CRM failure is viewing CRM as a technology initiative (Kale 2004). For this reason, we review the definitions in the Appendix with special attention to their emphasis on technology. This review suggests that CRM can be defined from at least three perspectives: narrowly and tactically as a particular technology solution, wide-ranging technology, and customer centric. These perspectives can be portrayed as a continuum (see Figure 1).

One organization we interviewed, which spent more than $30 million on IT solutions and systems integration, described CRM solely in terms of its SFA project. At this extreme, CRM is defined narrowly and tactically as a particular technology solution (e.g., Khanna 2001). We call this CRM "Perspective 1." Other definitions, such as that of Kutner and Cripps (1997), though somewhat broader, also fall into this category.

In another organization that we interviewed, the term CRM was used to refer to a wide range of customer-oriented IT and Internet solutions, reflecting Stone and Woodcock's (2001) definition. This represented CRM "Perspective 2," a point near the middle of the continuum.

"Perspective 3" reflects a more strategic and holistic approach to CRM that emphasizes the selective management of customer relationships to create shareholder value. This reflects elements of several previously noted definitions of CRM, including those of Buttle (2001), Glazer (1997), Singh and Agrawal (2003), and Swift (2000). Following this phase of our work, we identified Zablah, Beuenger, and Johnston's (2003) research, which supported our view of these perspectives.

The importance of how CRM is defined is not merely semantic. Its definition significantly affects the way an entire organization accepts and practices CRM. From a strategic viewpoint, CRM is not simply an IT solution that is used to acquire and grow a customer base; it involves a profound synthesis of strategic vision; a corporate understanding of the nature of customer value in a multichannel environment; the utilization of the appropriate information management and CRM applications; and high-quality operations, fulfillment, and service. Thus, we propose that in any organization, CRM should be positioned in the broad strategic context of Perspective 3.

Swift (2000) argues, and we concur, that organizations will benefit from adopting a relevant strategic CRM definition for their firm and ensuring its consistent use throughout their organization. Thus, we developed a definition of CRM that reflected Perspective 3. We examined the CRM literature, synthesized aspects of the various definitions into a draft definition, and then tested it with practicing managers. As our research progressed, we went through several iterations. The result is the following definition, which we use for the purposes of this study:

CRM is a strategic approach that is concerned with creating improved shareholder value through the development of appropriate relationships with key customers and customer segments. CRM unites the potential of relationship marketing strategies and IT to create profitable, long-term relationships with customers and other key stakeholders. CRM provides enhanced opportunities to use data and information to both understand customers and co-create value with them. This requires a cross-functional integration of processes, people, operations, and marketing capabilities that is enabled through information, technology, and applications.

This definition provided guidance for our subsequent research considerations and the strategic and cross-functional emphasis of the conceptual framework we developed.

Gartner (2001) calls for a fresh approach to business processes in CRM that involves both rethinking how these processes appear to the customer and reengineering them to be more customer centric. Kale (2004) supports this view and argues that a critical aspect of CRM involves identifying all strategic processes that take place between an enterprise and its customers. To address this challenge of adopting a fresh approach to CRM processes, we aimed to identify the key generic processes relevant to CRM.

We examined the literature to identify appropriate criteria for process selection but found little work in this area, with the exception of the contribution by Srivastava, Shervani, and Fahey (1999), who establish four process selection criteria for marketing and business processes. We chose their work as a starting point for the identification of process selection criteria for CRM. The criteria these authors propose are as follows: First, the processes should comprise a small set that addresses tasks critical to the achievement of an organization's goals. Second, each process should contribute to the value creation process. Third, each process should be at a strategic or macro level. Fourth, the processes need to manifest clear interrelationships.

As part of our research, we conducted a workshop with a panel of 34 highly experienced CRM practitioners, all of whom had extensive experience in the CRM and IT sectors. The director of a leading research and management institute specializing in the CRM and IT sectors selected the panel. Participants were selected on the basis of the following attributes to ensure that they were knowledgeable about CRM, its implementation, and its operation: substantial management and industrial experience (average of 17.2 years), maturity (average age of 40.2 years), international representation and international experience (managers from nine countries attended; most of them had international experience), and academic qualifications (degree or equivalent). In the first part of the workshop, which involved small group sessions, the panel reviewed and subsequently unanimously agreed that these four criteria were fully appropriate for selecting CRM processes. However, they also proposed two further criteria: First, each process should be cross-functional in nature, and second, each process would be considered by experienced practitioners as being both logical and beneficial to understanding and developing strategic CRM activities. We used these six criteria to select key generic CRM processes.

Grabner-Kraeuter and Moedritscher (2002) suggest that the absence of a strategic framework for CRM from which to define success is one reason for the disappointing results of many CRM initiatives. This view was supported both by the senior executives we interviewed during our research and by Gartner's (2001) research. Our next challenges were to identify key generic CRM processes using the previously described selection criteria and to develop them into a conceptual framework for CRM strategy development.

Our literature review found that few CRM frameworks exist; those that did were not based on a process-oriented cross-functional conceptualization of CRM. For example, Sue and Morin (2001, p. 6) outline a framework for CRM based on initiatives, expected results, and contributions, but this is not process based, and "many initiatives are not explicitly identified in the framework." Winer (2001, p. 91) develops a "basic model, which contains a set of 7 basic components: a database of customer activity; analyses of the database; given the analyses, decisions about which customers to target; tools for targeting the customers; how to build relationships with the targeted customers; privacy issues; and metrics for measuring the success of the CRM program." Again, this model, though useful, is not a cross-functional process-based conceptualization. This gap in the literature suggests that there is a need for a new systematic process-based CRM strategy framework. Synthesis of the diverse concepts in the literature on CRM and relationship marketing into a single, process-based framework should provide practical insights to help companies achieve greater success with CRM strategy development and implementation.

Conceptual frameworks and theory are typically based on combining previous literature, common sense, and experience (Eisenhardt 1989). In this research, we integrated a synthesis of the literature with learning from field-based interactions with executives to develop and refine the CRM strategy framework. In this approach, we used what Gummesson (2002a) terms "interaction research." This form of research originates from his view that "interaction and communication play a crucial role" in the stages of research and that testing concepts, ideas, and results through interaction with different target groups is "an integral part of the whole research process" (p. 345). The sources for these field-based insights, which include executives primarily from large enterprises in the business-to-business and business-to-consumer sectors, included the following:

• An expert panel of 34 highly experienced executives;

• Interviews with 20 executives working in CRM, marketing, and IT roles in companies in the financial services sector;

• Interviews with six executives from large CRM vendors and with five executives from three CRM and strategy consultancies;

• Individual and group discussions with CRM, marketing, and IT managers at workshops with 18 CRM vendors, analysts, and their clients, including Accenture, Baan, BroadVision, Chordiant, EDS, E.piphany, Hewlett-Packard, IBM, Gartner, NCR Teradata, Peoplesoft, Oracle, SAP, SAS Institute, Siebel, Sybase, and Unisys;

• Piloting the framework as a planning tool in the financial services and automotive sectors; and

• Using the framework as a planning tool in two companies: global telecommunications and global logistics. Six workshops were held in each company.

We began by identifying possible generic CRM processes from the CRM and related business literature. We then discussed these tentative processes interactively with the groups of executives. The outcome of this work was a short list of seven processes. We then used the expert panel of experienced CRM executives who had assisted in the development of the process selection schema to nominate the CRM processes that they considered important and to agree on those that were the most relevant and generic. After an initial group workshop, each panel member independently completed a list representing his or her view of the key generic processes that met the six previously agreed-on process criteria. The data were fed back to this group, and a detailed discussion followed to help confirm our understanding of the process categories.

As a result of this interactive method, five CRM processes that met the selection criteria were identified; all five were agreed on as important generic processes by more than two-thirds of the group in the first iteration. Subsequently, we received strong confirmation of these as key generic CRM processes by several of the other groups of managers. The resultant five generic processes were ( 1) the strategy development process, ( 2) the value creation process, ( 3) the multichannel integration process, ( 4) the information management process, and ( 5) the performance assessment process.

We then incorporated these five key generic CRM processes into a preliminary conceptual framework. This initial framework and the development of subsequent versions were both informed by and further refined by our interactions with two primary executive groups: mangers from the previously noted companies and executives from three CRM consulting firms. Participants at several academic conferences on CRM and relationship marketing also assisted with comments and criticisms of previous versions. With evolving versions of the framework, we combined a synthesis of relevant literature with field-based interactions involving the groups. The framework went through a considerable number of major iterations and minor revisions; the final version appears in Figure 2.

This conceptual framework illustrates the interactive set of strategic processes that commences with a detailed review of an organization's strategy (the strategy development process) and concludes with an improvement in business results and increased share value (the performance assessment process). The concept that competitive advantage stems from the creation of value for the customer and for the business and associated cocreation activities (the value creation process) is well developed in the marketing literature. For large companies, CRM activity will involve collecting and intelligently using customer and other relevant data (the information process) to build a consistently superior customer experience and enduring customer relationships (the multichannel integration process). The iterative nature of CRM strategy development is highlighted by the arrows between the processes in both directions in Figure 2; they represent interaction and feedback loops between the different processes. The circular arrows in the value creation process reflect the cocreation process. We now examine the key components we identified in each process. As with our prior work, we used the interaction research method in the identification of these process components.

This process requires a dual focus on the organization's business strategy and its customer strategy. How well the two interrelate fundamentally affects the success of its CRM strategy.

The business strategy must be considered first to determine how the customer strategy should be developed and how it should evolve over time. The business strategy process can commence with a review or articulation of a company's vision, especially as it relates to CRM (e.g., Davidson 2002). Next, the industry and competitive environment should be reviewed. Traditional industry analysis (e.g., Porter 1980) should be augmented by more contemporary approaches (e.g., Christensen 2001; Slater and Olson 2002) to include co-opetition (Brandenburger and Nalebuff 1997), networks and deeper environmental analysis (Achrol 1997), and the impact of disruptive technologies (Christensen and Overdorf 2000).

Whereas business strategy is usually the responsibility of the chief executive officer, the board, and the strategy director, customer strategy is typically the responsibility of the marketing department. Although CRM requires a cross-functional approach, it is often vested in functionally based roles, including IT and marketing. When different departments are involved in the two areas of strategy development, special emphasis should be placed on the alignment and integration of business strategy.

Customer strategy involves examining the existing and potential customer base and identifying which forms of segmentation are most appropriate. As part of this process, the organization needs to consider the level of subdivision for customer segments, or segment granularity. This involves decisions about whether a macro, micro, or one-to-one segmentation approach is appropriate (Rubin 1997).

Several authors emphasize the potential for shifting from a mass market to an individualized, or one-to-one, marketing environment. Exploiting e-commerce opportunities and the fundamental economic characteristics of the Internet can enable a much deeper level of segmentation granularity than is affordable in most other channels (e.g., Peppers and Rogers 1993, 1997). In summary, the strategy development process involves a detailed assessment of business strategy and the development of an appropriate customer strategy. This should provide the enterprise with a clearer platform on which to develop and implement its CRM activities.

The value creation process transforms the outputs of the strategy development process into programs that both extract and deliver value. The three key elements of the value creation process are ( 1) determining what value the company can provide to its customer; ( 2) determining what value the company can receives from its customers; and ( 3) by successfully managing this value exchange, which involves a process of cocreation or coproduction, maximizing the lifetime value of desirable customer segments.

The value the customer receives from the organization draws on the concept of the benefits that enhance the customer offer (Levitt 1969; Lovelock 1995). However, there is now a logic, which has evolved from earlier thinking in business-to-business and services marketing, that views the customer as a cocreator and coproducer (Bendapudi and Leone 2003; Prahalad and Ramaswamy 2004; Vargo and Lusch 2004). These benefits can be integrated in the form of a value proposition (e.g., Lanning and Michaels 1988; Lanning and Phillips 1991) that explains the relationship among the performance of the product, the fulfillment of the customer's needs, and the total cost to the customer over the customer relationship life cycle (Lanning and Michaels 1988). Lanning's (1998) later work on value propositions reflects the cocreation perspective. However, a more detailed synthesis of work in this area is needed in further research.

To determine whether the value proposition is likely to result in a superior customer experience, a company should undertake a value assessment to quantify the relative importance that customers place on the various attributes of a product. Analytical tools such as conjoint analysis can be used to identify customers that share common preferences in terms of product attributes. Such tools may also reveal substantial market segments with service needs that are not fully catered to by the attributes of existing offers.

From this perspective, customer value is the outcome of the coproduction of value, the deployment of improved acquisition and retention strategies, and the utilization of effective channel management. Fundamental to this concept of customer value are two key elements that require further research. First, it is necessary to determine how existing and potential customer profitability varies across different customers and customer segments. Second, the economics of customer acquisition and customer retention and opportunities for cross-selling, up-selling, and building customer advocacy must be understood. How these elements contribute to increasing customer lifetime value is integral to value creation.

Customer retention represents a significant part of the research on value creation. For example, Reichheld and Sasser (1990) identify the net present value profit improvement of retaining customers, and Rust and Zahorik (1993) and Rust, Zahorik, and Keiningham (1995) outline procedures for assessing the impact of satisfaction and quality improvement efforts on customer retention and market share. More recently, research has emphasized customer equity (e.g., Blattberg and Deighton 1996; Hogan, Lemon, and Rust 2002; Rust, Lemon, and Zeithaml 2004). Calculating the customer lifetime value of different segments enables organizations to focus on the most profitable customers and customer segments. The value creation process is a crucial component of CRM because it translates business and customer strategies into specific value proposition statements that demonstrate what value is to be delivered to customers, and thus, it explains what value is to be received by the organization, including the potential for cocreation.

The multichannel integration process is arguably one of the most important processes in CRM because it takes the outputs of the business strategy and value creation processes and translates them into value-adding activities with customers. However, there is only a small amount of published work on the multichannel integration in CRM (e.g., Friedman and Furey 1999; Funk 2002; Kraft 2000; Sudharshan and Sanchez 1998; Wagner 2000). The multichannel integration process focuses on decisions about what the most appropriate combinations of channels to use are; how to ensure that the customer experiences highly positive interactions within those channels; and when a customer interacts with more than one channel, how to create and present a single unified view of the customer.

Today, many companies enter the market through a hybrid channel model (Friedman and Furey 1999; Moriarty and Moran 1990) that involves multiple channels, such as field sales forces, Internet, direct mail, business partners, and telephony. There are a growing number of channels by which a company can interact with its customers. Through an iterative process, we categorized the many channel options into six categories broadly based on the balance of physical or virtual contact (see Figure 2). These include ( 1) sales force, including field account management, service, and personal representation; ( 2) outlets, including retail branches, stores, depots, and kiosks; ( 3) telephony, including traditional telephone, facsimile, telex, and call center contact; ( 4) direct marketing, including direct mail, radio, and traditional television (but excluding e-commerce); ( 5) e-commerce, including e-mail, the Internet, and interactive digital television; and ( 6) m-commerce, including mobile telephony, short message service and text messaging, wireless application protocol, and 3G mobile services. Some channels are now being used in combination to maximize commercial exposure and return; for example, there is collaborative browsing and Internet relay chat, used by companies such as Lands End, and voice over IP (Internet protocol), which integrates both telephony and the Internet.

Managing integrated channels relies on the ability to uphold the same high standards across multiple, different channels. Having established a set of standards for each channel that defines an outstanding customer experience for that channel, the organization can then work to integrate the channels. The concept of the "perfect customer experience," which must be affordable for the company in the context of the segments in which it operates and its competition, is a relatively new concept. This concept is now being embraced in industry by companies such as TNT, Toyota's Lexus, Oce, and Guinness Breweries, but it has yet to receive much attention in the academic literature. Therefore, multichannel integration is a critical process in CRM because it represents the point of cocreation of customer value. However, a company's ability to execute multichannel integration successfully is heavily dependent on the organization's ability to gather and deploy customer information from all channels and to integrate it with other relevant information.

The information management process is concerned with the collection, collation, and use of customer data and information from all customer contact points to generate customer insight and appropriate marketing responses. The key material elements of the information management process are the data repository, which provides a corporate memory of customers; IT systems, which include the organization's computer hardware, software, and middleware; analysis tools; and front office and back office applications, which support the many activities involved in interfacing directly with customers and managing internal operations, administration, and supplier relationships (Greenberg 2001).

The data repository provides a powerful corporate memory of customers, an integrated enterprisewide data store that is capable of relevant data analyses. In larger organizations, it may comprise a data warehouse (Agosta 1999; Swift 2000) and related data marts and databases. There are two forms of data warehouse, the conventional data warehouse and the operational data store. The latter stores only the information necessary to provide a single identity for all customers. An enterprise data model is used to manage this data conversion process to minimize data duplication and to resolve any inconsistencies between databases.

Information technology systems refer to the computer hardware and the related software and middleware used in the organization. Often, technology integration is required before databases can be integrated into a data warehouse and user access can be provided across the company. However, the historical separation between marketing and IT sometimes presents integration issues at the organizational level (Glazer 1997). The organization's capacity to scale existing systems or to plan for the migration to larger systems without disrupting business operations is critical.

The analytical tools that enable effective use of the data warehouse can be found in general data-mining packages and in specific software application packages. Data mining enables the analysis of large quantities of data to discover meaningful patterns and relationships (e.g., Groth 2000; Peacock 1998). More specific software application packages include analytical tools that focus on such tasks as campaign management analysis, credit scoring, and customer profiling.

Front office applications are the technologies a company uses to support all those activities that involve direct interface with customers, including SFA and call center management. Back office applications support internal administration activities and supplier relationships, including human resources, procurement, warehouse management, logistics software, and some financial processes. A key concern about the front and back office systems offered by CRM vendors is that they are sufficiently connected and cocoordinated to improve customer relationships and workflow.

Gartner segments vendors of CRM applications and CRM service providers into specific categories (Radcliffe and Kirkby 2002), and Greenberg (2001) and Jacobsen (1999) provide detailed reviews of CRM vendors' products. The key segments for CRM applications are Integrated CRM and Enterprise Resource Planning Suite (e.g., Oracle, PeopleSoft, SAP), CRM Suite (e.g., E.piphany, Siebel), CRM Framework (e.g., Chordiant), CRM Best of Breed (e.g., NCR Teradata; Broadvision), and "Build it Yourself" (e.g., IBM, Oracle, Sun). The CRM service providers and consultants that offer implementation support specialize in the following areas: corporate strategy (e.g., McKinsey, Bain); CRM strategy (e.g., Peppers & Rogers, Vectia); change management, organization design, training, human resources, and so forth (e.g., Accenture); business transformation (e.g., IBM); infrastructure building and systems integration (e.g., Siemens, Unisys); infrastructure outsourcing (e.g., EDS, CSC); business insight, research, and so forth (e.g., SAS); and business process outsourcing (e.g., Acxiom). The need for comprehensive and scalable options has created scope for many new products from CRM vendors. However, despite their claim to be "complete CRM solution providers," few software vendors can provide the full range of functionality that a complete CRM business strategy requires.

The information management process provides a means of sharing relevant customer and other information throughout the enterprise and "replicating the mind of the customer." To ensure that technology solutions support CRM, it is important to conduct IT planning from a perspective of providing a seamless customer service rather than planning for functional or product-centered departments and activities. Furthermore, data analysis tools should measure business activities. This kind of analysis provides the basis for the performance assessment process.

The performance assessment process covers the essential task of ensuring that the organization's strategic aims in terms of CRM are being delivered to an appropriate and acceptable standard and that a basis for future improvement is established. This process can be viewed as having two main components: shareholder results, which provide a macro view of the overall relationships that drive performance, and performance monitoring, which provides a more detailed, micro view of metrics and key performance indicators.

To achieve the ultimate objective of CRM, the delivery of shareholder results, the organization should consider how to build employee value, customer value, and shareholder value and how to reduce costs. Recent research on relationships among employees, customers, and shareholders has emphasized the need to adopt a more informed and integrated approach to exploiting the linkages among them. The service profit chain model and related research focuses on establishing the relationships among employee satisfaction, customer loyalty, profitability, and shareholder value (e.g., Heskett et al. 1994; Loveman 1998). Organizations also need to focus on cost reduction opportunities. Two means of cost reduction are especially relevant to CRM: deployment of technologies ranging from automated telephony services to Web services and the use of new electronic channels such as online, self-service facilities. The development of models such as the service profit chain has been important in enabling companies to consider the effectiveness of CRM at a strategic level in terms of improving shareholder results.

Despite a growing call for companies to be more customer oriented, there is concern that, in general, the metrics used by companies to measure and monitor their CRM performance are not well developed or well communicated. Ambler's (2002) research findings raise particular concern; he finds that key aspects of CRM, such as customer satisfaction and customer retention, only reach the board in 36% and 51% of companies, respectively. Even when these metrics reach the board level, it is not clear how deeply they are understood and how much time is spent on them. Traditional performance measurement systems, which tend to be functionally driven, may be inappropriate for cross-functional CRM.

Recent efforts to provide cross-functional measures, such as the balanced scorecard (Kaplan and Norton 1996), are a useful advance. The format of the balanced scorecard enables a wide range of metrics designs. Indicators that can reveal future financial results, not just historical results, need to be considered as part of this process. Standards, metrics, and key performance indicators for CRM should reflect the performance standards necessary across the five major processes to ensure that CRM activities are planned and practiced effectively and that a feedback loop exists to maximize performance improvement and organizational learning. A consideration of "return on relationships" (Gummesson 2004) will assist in identifying further metrics that are relevant to the enterprise.

In this article, we develop a cross-functional, process-based CRM strategy framework that aims to help companies avoid the potential problems associated with a narrow technological definition of CRM and realize strategic benefits. Our research was based on large industrial companies because the size and complexity of such enterprises is likely to present the greatest CRM challenges. We did not examine issues related to small or medium-sized companies and nonprofit organizations in this work.

This study contributes to the marketing literature in several ways. First, our work extends a managerial perspective that stresses the importance of cross-functional processes in CRM strategy and contributes to the positioning of the poorly defined CRM concept within the marketing literature. Second, it provides a process-based conceptual framework for strategic CRM and identifies key elements within each process. Third, it makes a contribution to the limited literature on interaction research. Finally, the research represents a grounded contribution that offers managers insight into the development and implementation of CRM strategies. To date, this framework has been used by companies to address several issues, including surfacing problematic CRM issues, planning the key components of a CRM strategy, identifying which process components of CRM should receive priority, creating a platform for change, and benchmarking other companies' CRM activities.

Much research remains to be done in the exploration of the multifaceted nature of CRM. Sheth (1996) notes that for an emerging management discipline, it is important to have an acceptable definition that encompasses all facets to focus understanding and growth of knowledge in the discipline. He proposes a multistage process for achieving this that begins with delimiting the domain, agreeing on a definition, developing performance measures, and developing explanatory theory. The framework we propose in this article offers a potentially useful starting point for the development of improved insight into these aspects of CRM theory. The task of delimiting the domain, agreeing on a definition for CRM, and building a research agenda will be an evolving process in this nascent area. We do not attempt to build such a research agenda in the current work; however, we emphasize the importance of CRM implementation and related people issues as an area in which further research is urgently needed. Initial work by Ebner and colleagues (2002), Gummesson (2002b, c), Henneberg (2003), Pettit (2002), and Rigby, Reichheld, and Schefter (2002) provides a useful platform from which to develop this important research area.

The authors acknowledge the financial support of BT plc and SAS with this research, and they thank the three anonymous JM reviewers and the consulting editors for their helpful comments on previous versions of this article.

CRM Defined Narrowly and Tactically

CRM Defined Broadly and Strategically

CRM is about the implementation of a specific technology solution project.

CRM is the implementation of an integrated series of customer-oriented technology solutions.

CRM is a holistic approach to managing customer relationships to create shareholder value.

DIAGRAM: FIGURE 2; A Conceptual Framework for CRM Strategy

Achrol, Ravi S. (1997), "Changes in the Theory of Interorganizational Relations in Marketing: Toward a Network Paradigm," Journal of the Academy of Marketing Science, 25 (Winter), 56-71.

Agosta, Lou (1999), The Essential Guide to Data Warehousing. Upper Saddle River, NJ: Prentice Hall.

Ambler, Tim (2002), Marketing and the Bottom Line, 2d ed. London: Financial Times/Prentice Hall.

Bendapudi, Neeli and Robert P. Leone (2003), "Psychological Implications of Customer Participation in Co-Production," Journal of Marketing, 67 (January), 14-28.

Blattberg, Robert C. and John Deighton (1996), "Managing Marketing by the Customer Equity Criterion," Harvard Business Review, 74 (July-August), 136-44.

Brandenburger, Adam M. and Barry J. Nalebuff (1997), Coopetition: A Revolution Mindset That Combines Competition and Game Theory Strategy That's Changing the Game of Business. New York: Doubleday.

Buttle, Francis A. (2001), "The CRM Value Chain," Marketing Business, (February), 52-55.

Christensen, Clayton M. (2001), "The Past and Future of Competitive Advantage," Sloan Management Review, 42 (Winter), 105-109.

----- and Michael Overdorf (2000), "Meeting the Challenge of Disruptive Change," Harvard Business Review, 78 (March-April), 67-76.

Couldwell, Clive (1999), "Loyalty Bonuses," Marketing Week, (February 18), 14.

Davidson, Hugh (2002), The Committed Enterprise. Oxford: Butterworth-Heinemann.

Ebner, Manuel, Arthur Hu, Daniel Levitt, and Jim McCory (2002), "How to Rescue CRM," McKinsey Quarterly, (Special Edition), 49-57.

Eisenhardt, Kathleen M. (1989), "Building Theories from Case Study Research," Academy of Management Review, 14 (March), 532-50.

Friedman, Lawrence G. and Tim R. Furey (1999), The Channel Advantage. Oxford: Butterworth-Heinemann.

Funk, Jeffrey L. (2002), "Mobile Shopping and Multi-Channel Integration in the Japanese Mobile Industry," working paper, Research Institute for Economics and Business Administration, Kobe University.

Gartner Group (2001), "CRM Economics: Figuring Out the ROI on Customer Initiatives," white paper, Stamford, CT.

Glazer, Rashi (1997), "Strategy and Structure in Information-Intensive Markets: The Relationship Between Marketing and IT," Journal of Market Focused Management, 2 (1), 65-81.

Gosney, John and Thomas Boehm (2000), Customer Relationship Management Essentials. Indianapolis, IN: Prima Publishing.

Grabner-Kraeuter, Sonja and Gernot Moedritscher (2002), "Alternative Approaches Toward Measuring CRM Performance," paper presented at the Sixth Research Conference on Relationship Marketing and Customer Relationship Management, Atlanta (June 9-12).

Greenberg, Paul (2001), CRM at the Speed of Light. Berkeley, CA: Osborne/McGraw-Hill.

Groth, Robert (2000), Data Mining: Building Competitive Advantage. Upper Saddle River, NJ: Prentice Hall.

Gummesson, Evert (2002a), "Practical Value of Adequate Marketing Management Theory," European Journal of Marketing, 36 (March), 325-49.

----- (2002b), "Relationship Marketing in the New Economy," Journal of Relationship Marketing, 1 (1), 37-57.

----- (2002c), Total Relationship Marketing, 2d ed. Oxford: Butterworth-Heinemann.

----- (2004), "Return on Relationships (ROR): The Value of Relationship Marketing and CRM in Business-to-Business Contexts," Journal of Business and Industrial Marketing, 1 (2), 136-48.

Henneberg, Stephan C.M. (2003), "An Exploratory Analysis of CRM Implementation Models," paper presented at the 11th International Colloquium in Relationship Marketing, University of Gloucestershire (September).

Heskett, James L., Thomas O. Jones, Gary W. Loveman, Earl W. Sasser Jr., and Leonard A. Schlesinger (1994), "Putting the Service-Profit Chain to Work," Harvard Business Review, 82 (March-April), 164-74.

Hobby, John (1999), "Looking After the One Who Matters," Accountancy Age, (October 28), 28-30.

Hogan, John E., Katherine N. Lemon, and Roland T. Rust (2002), "Customer Equity Management: Charting New Directions for the Future of Marketing," Journal of Service Research, 5 (August), 4-13.

Jacobsen, Per Ostergaard, ed. (1999), The CRM Handbook from Group to Multiindividual. Hellerup, Denmark: PricewaterhouseCoopers.

Kale, Sudhir H. (2004), "CRM Failure and the Seven Deadly Sins," Marketing Management, 13 (September-October), 42-46.

Kaplan, Robert S. and David P. Norton (1996), The Balanced Scorecard: Translating Strategy into Action. Boston: Harvard Business School Press.

Khanna, Sunil (2001), "Measuring the CRM ROI: Show Them Benefits," (accessed November 20, 2002), [available at http:// www.crm-forum.com].

Kraft, Kevin J. (2000), "Multi-Channel Customer Management Solutions: Making Sense of the Operational Issues and Deployment Alternatives," white paper, eLoyalty Corporation, Chicago.

Kutner, Steve and John Cripps (1997), "Managing the Customer Portfolio of Healthcare Enterprises," The Healthcare Forum Journal, 4 (September-October), 52-54.

Lanning, Michael J. (1998), Delivering Profitable Value. Cambridge, MA: Perseus Publishing.

----- and Edward G. Michaels (1988), "A Business Is a Value Delivery System," staff paper, McKinsey & Company.

----- and Lynn Phillips (1991), "Building Market-Focused Organizations," white paper, Gemini Consulting.

Levitt, Theodore (1969), The Marketing Mode: Pathways to Corporate Growth. New York: McGraw-Hill.

Lovelock, Christopher (1995), "Competing on Service: Technology and Teamwork in Supplementary Services," Planning Review, 23 (July-August), 32-39.

Loveman, Gary W. (1998), "Employee Satisfaction, Customer Loyalty and Financial Performance," Journal of Service Research, 1 (August), 18-31.

Moriarty, Rowland T. and Ursula Moran (1990), "Managing Hybrid Marketing Systems," Harvard Business Review, 68 (November-December), 146-57.

Parvatiyar, Atul and Jagdish N. Sheth (2001), "Conceptual Framework of Customer Relationship Management," in Customer Relationship Management--Emerging Concepts, Tools and Applications, Jagdish N. Sheth, Atul Parvatiyar, and G. Shainesh, eds. New Delhi, India: Tata/McGraw-Hill, 3-25.

Peacock, Peter R. (1998), "Data Mining in Marketing: Part 1," Marketing Management, 7 (Winter), 9-18.

Peppers, Don and Martha Rogers (1993), The One to One Future. London: Piatkus.

----- and ----- (1997), Enterprise One to One. London: Piatkus.

-----, -----, and Bob Dorf (1999), "Is Your Company Ready for One-to-One Marketing?" Harvard Business Review, 77 (January-February), 151-60.

Pettit, Raymond (2002), "The State of CRM: Addressing Efficiencies and the Achilles' Heel of CRM," Business Intelligence Advisory Service executive report, 2 (3).

Porter, Michael E. (1980), Competitive Strategy. New York: The Free Press.

Prahalad, C.K. and Venkatram Ramaswamy (2004), "Co-creating Unique Value with Customers," Strategy and Leadership, 32 (3), 4-9.

Radcliffe, John and Jennifer Kirkby (2002), "CRM Vendors and Service Providers," paper presented at the Gartner Symposium ITXPO, Florence, Italy (April 8-10).

Reichheld, Frederick F. and W. Earl Sasser Jr. (1990), "Zero Defections: Quality Comes to Services," Harvard Business Review, 68 (September-October), 105-111.

Reinartz, Wemer, Manfred Krafft, and Wayne D. Hoyer (2004), "The Customer Relationship Management Process: Its Measurement and Impact on Performance," Journal of Marketing Research, 41 (August), 293-305.

Rigby, Darrell K., Frederick F. Reichheld, and Phil Schefter (2002), "Avoid the Four Perils of CRM," Harvard Business Review, 80 (February), 101-109.

Rubin, Marc (1997), "Creating Customer-Oriented Companies," Prism, 4 (4), 5-27.

Rust, Roland T., Katherine N. Lemon, and Valarie A. Zeithaml (2004), "Return on Marketing: Using Customer Equity to Focus Marketing Strategy," Journal of Marketing, 68 (January), 109-128.

----- and Anthony J. Zahorik (1993), "Customer Satisfaction, Customer Retention and Market Share," Journal of Retailing, 69 (Summer) 193-215.

-----, -----, and Timothy L. Keiningham (1995), "Return on Quality (ROQ): Making Service Quality Financially Accountable," Journal of Marketing, 59 (April), 58-70.

Ryals, Lynette and Adrian F.T. Payne (2001), "Customer Relationship Management in Financial Services: Towards Information-Enabled Relationship Marketing," Journal of Strategic Marketing, 9 (March), 1-25.

Sheth, Jagdish N. (1996), "Relationship Marketing: Paradigm Shift or Shaft?" paper presented at the Annual Meeting of the Academy of Marketing Science, Miami (May).

Singh, Deepali and D.P Agrawal (2003), "CRM Practices in Indian Industries," International Journal of Customer Relationship Management, 5 (December-January), 241-57.

Slater, Stanley F. and Eric M. Olson (2002), "A Fresh Look at Industry and Market Analysis," Business Horizons, 45 (January-February), 15-22.

Srivastava, Rajendra K., Tasadduq A. Shervani, and Liam Fahey (1999), "Marketing, Business Processes, and Shareholder Value: An Organizationally Embedded View of Marketing Activities and the Discipline of Marketing," Journal of Marketing, 63 (Special Issue), 168-79.

Stone, Merlin and Neil Woodcock (2001), "Defining CRM and Assessing its Quality," in Successful Customer Relationship Marketing, Brian Foss and Merlin Stone, eds. London: Kogan Page, 3-20.

Sudharshan, D. and Ron Sanchez (1998), "Distribution Equity: Creating Value Through Managing Knowledge Relationships with Distribution Channels," Journal of Market Focused Management, 2 (4), 309-338.

Sue, Patrick and Paul Morin (2001), "A Strategic Framework for CRM," (accessed April 13, 2002), [available at http://www. crm-forum.com].

Swift, Ronald S. (2000), Accelerating Customer Relationships--Using CRM and Relationship Technologies. Upper Saddle River, NJ: Prentice Hall.

Vargo Stephen L. and Robert R. Lusch (2004), "Evolving to a New Dominant Logic for Marketing," Journal of Marketing, 68 (January), 1-17.

Wagner, Todd R. (2000), "Multichannel Customer Interaction," in Defying the Limits. San Francisco: Montgomery Research, 277-80.

Winer, Russell S. (2001), "A Framework for Customer Relationship Management," California Management Review, 43 (Summer), 89-105.

Zablah, Alex R., Danny N. Beuenger, and Wesley J. Johnston (2003), "Customer Relationship Management: An Explication of Its Domain and Avenues for Further Inquiry," in Relationship Marketing, Customer Relationship Management and Marketing Management: Co-Operation-Competition-Co-Evolution, Michael Kleinaltenkamp and Michael Ehret, eds. Berlin: Freie Universität Berlin, 115-24.

• CRM is an e-commerce application (Khanna 2001).

• CRM is a term for methodologies, technologies, and e-commerce capabilities used by companies to manage customer relationships (Stone and Woodcock 2001).

• CRM is an enterprisewide initiative that belongs in all areas of an organization (Singh and Agrawal 2003).

• CRM is a comprehensive strategy and process of acquiring, retaining, and partnering with selective customers to create superior value for the company and the customer (Parvitiyar and Sheth 2001).

• CRM is about the development and maintenance of long-term, mutually beneficial relationships with strategically significant customers (Buttle 2001).

• CRM includes numerous aspects, but the basic theme is for the company to become more customer-centric. Methods are primarily Web-based tools and Internet presence (Gosney and Boehm 2000).

• CRM can be viewed as an application of one-to-one marketing and relationship marketing, responding to an individual customer on the basis of what the customer says and what else is known about that customer (Peppers, Rogers, and Dorf 1999).

• CRM is a management approach that enables organizations to identify, attract, and increase retention of profitable customers by managing relationships with them (Hobby 1999).

• CRM involves using existing customer information to improve company profitability and customer service (Couldwell 1999).

• CRM attempts to provide a strategic bridge between information technology and marketing strategies aimed at building long-term relationships and profitability. This requires "information-intensive strategies" (Glazer 1997).

• CRM is data-driven marketing (Kutner and Cripps 1997).

• CRM is an enterprise approach to understanding and influencing customer behavior through meaningful communication to improve customer acquisition, customer retention, customer loyalty, and customer profitability (Swift 2000).

~~~~~~~~

By Adrian Payne and Pennie Frow

Adrian Payne is Professor of Services and Relationship Marketing and Director of the Centre for CRM, Cranfield School of Management, Cranfield University

Pennie Frow is Visiting Fellow in Marketing, Cranfield School of Management, Cranfield University

Actualizing Innovation Effort: The Impact of Market Knowledge Diffusion in a Dynamic System of Competition 

This study focuses on the dynamic process that governs the impact of market knowledge diffusion on innovation effort and its subsequent effect on firm performance. First, the author proposes that three aspects of market knowledge (knowledge level, knowledge change, and extent of shared knowledge about customers and competitors) influence innovation effort. In so doing, she explicitly models the dynamic process of competition, including heterogeneity in interdependence of innovation across competitors, firm-specific inertial tendency in innovation, and feedback effects reflected in satisfaction with past performance. Second, within a partial adjustment model of performance, the author studies the role of shared market knowledge and firm size in the translation of innovation effort into firm performance over time. She tests the conceptual framework with longitudinal quasi field experiments based on a Markstrat simulation exercise, including a main experiment and three validation studies. The results reveal a dynamic system in which some aspects of market knowledge diffusion propel innovation, whereas satisfaction with past performance hinders innovation effort. Furthermore, the results show that innovation effort, by itself, does not affect firm performance. In the context of the study, total shared market knowledge helps smaller firms actualize better returns from their innovation effort than larger firms.

At a fundamental level, firms act on the basis of their market knowledge: their knowledge of customers and competitors. As a result, the concepts of knowledge sharing (e.g., Hoopes and Postrel 1999; Senge 1993) and organizational learning (e.g., Cyert and March 1963; Dickson 1992) have gained substantial attention from scholars and business practitioners (e.g., Gates 1999). In general, an organization's ability to recognize the value of new information, assimilate it, and use it strategically is regarded as crucial for its ability to innovate (see Cohen and Levinthal 1990) and to gain performance advantages (Day 1997). In practice, organizations implement a team approach in developing new products to ensure organizationwide knowledge acquisition and dissemination. It is expected that information that resides in isolated pockets in the organization will become shared over time, thus leading to better decision making.

The objective of this article is to investigate the dynamic process that governs the impact of market knowledge diffusion on innovation effort and subsequent firm performance. The article also aims to close three gaps that are currently evident in research on market knowledge and innovation. First, as do related organizational processes, innovation evolves over time and thus requires the use of knowledge in a dynamic setting. However, empirical research in several disciplines--including marketing, management, and organizational behavior--has largely been guided by cross-sectional analyses of a key informant's perceptions of strategic orientation, innovation, and performance. That is, there has been no systematic conceptualization or longitudinal assessment of the dynamic process of market knowledge diffusion and its impact on innovation and subsequent performance. Lee (2003) notes that success in innovation-driven strategic divergence can be achieved by the capture of internal and external knowledge spillovers, but no conceptual or empirical work has explicated this process.

Second, market knowledge diffusion is a function of information acquisition, which is shaped over time by the dynamics of product diffusion and firms' previous innovative activities. Although prior research recognizes general time dependence in performance outcomes (e.g., Boulding 1990; Jacobson 1990), few studies have accounted for firm-specific inertial tendency in innovation (e.g., Bayus, Erickson, and Jacobson 2003). I suggest that to diagnose how a firm's own market knowledge and strategic actions uniquely drive its innovation, it is necessary to account for firm-specific inertial tendencies over time. Ignoring of firm-specific inertial tendencies can lead to erroneous conclusions regarding the impact of market knowledge diffusion on innovation and subsequent performance.

Third, prior research has failed to consider the implications of a competitive system in which the actions of a firm's competitors influence the firm's strategic actions, as the dynamic use of market knowledge. Competitors often try to emulate the performance of successful others (Dickson 1992), but not all firms are equally good at identifying the explanatory mechanisms that underlie success. Two firms that follow similar strategies in using and managing market knowledge are likely to experience greater interdependence than are two firms that follow different strategies. To understand how a firm's own market knowledge and strategic actions drive its innovation or performance over time, it is necessary to account for the interdependence in actions and outcomes across competitors.

In an important departure from the extant literature, I adopt a longitudinal perspective that examines the evolution of market knowledge, innovation, and performance over time. I conceptualize market knowledge diffusion as the dynamics of market knowledge level, change in market knowledge, and shared market knowledge among strategic decision makers, and I subsequently develop and test hypotheses about these constructs impact on innovation effort (Figure 1 depicts the tested relationships). I isolate the impact of market knowledge diffusion within the firm by accounting for innovation interdependence among firms, firm-specific proneness to inertia in innovation, and market feedback effects. I then develop a dynamic model of firm performance that reveals the unique role of decision makers shared market knowledge in translating innovation effort into performance. Finally, to test the conceptual framework, I conduct longitudinal quasi field experiments in the context of the Markstrat simulation.

In addition to extending research in marketing, this study also has managerial implications. It calls attention to the role and relative importance of different aspects of market knowledge diffusion in strategic decision-making teams. Although many organizations are attempting to maximize their market knowledge (U.S. firms spend approximately $6 billion per year on market research information, according to the American Marketing Association), others are questioning its payoffs and impact on the bottom line (Sutcliffe and Weber 2003). Sutcliffe and Weber (2003) suggest that accurate market knowledge actually hurts performance and advise against reliance on it. The present research contributes to this debate by postulating three distinct aspects of market knowledge diffusion and by demonstrating the roles they play in generating performance.

Researchers have long suggested that markets operate to disseminate information rapidly to interested parties (von Hayek 1989). However, not all new market information becomes part of the intraorganizational knowledge-diffusion process. I suggest that it is the very mechanism of intraorganizational knowledge diffusion that produces the observed stop-and-go, or interrupted, pattern of innovation generation (Chandrashekaran et al. 1999; Dickson 1996; Hunt and Morgan 1995). If markets are constantly changing (Dickson 1992; Hunt and Morgan 1995) and if innovation drives change and firm success (Schumpeter 1942), then market knowledge, change in market knowledge, and shared market knowledge should enhance innovation effort. By adopting the premise that knowledge ultimately resides at the level of the individual strategic decision maker, this research offers a microlevel perspective on market knowledge diffusion. The premise aligns with Cyert and March's (1963) theoretical perspective of organizations as comprising individual members who learn and share knowledge with others and whose knowledge drives the specific investments of strategic decision-making teams. The premise is also consistent with the notion that market knowledge is information-based knowledge (Polanyi 1967) that resides with the individual decision maker (Spender 1996).

Definitions

In defining knowledge, and specifically market knowledge, I draw on the work of Hammond and Summers (1972) and maintain that knowledge reflects the extent of a subject's accurate detection of a task's properties. A decision maker who can correctly identify customer preferences and competitors is deemed knowledgeable about customer preferences and competitors. To specify the domain of market knowledge, I build on the work of Day and Nedungadi (1994, p. 32), who note that "the two most salient features of a competitive market are customers and competitors." Market factors such as competition and demand have been recognized as crucial determinants of organizational phenomena such as innovation generation (e.g., Kessler and Chakrabarti 1996; Porter 1995). At a fundamental level, knowledge of the market is necessary to determine the needs and wants of target markets and to satisfy them better than the competition can (Kotler 2000). Therefore, in this research, the term "market knowledge" implies knowledge about customers and competitors.

Change in market knowledge refers to the magnitude of change in decision makers' knowledge about customers and competitors between two points in time. This definition focuses on the absolute change in knowledge over time rather than the direction of change, which makes it possible to separate the effects of knowledge accuracy (captured by market knowledge, as defined previously) from the effects of the magnitude of change in market knowledge. Change in market knowledge may be viewed as a consequence of adjustment to inaccuracies of knowledge compared with the "objective" reality in the market.. Furthermore, this change may be due to insights beyond current available information and based on a decision maker's anticipating the market or uniquely integrating knowledge across different market aspects. Decision makers' desire to make good decisions motivates them to update and change their knowledge about the marketplace, but inertial forces that result from the cyclical nature of the market evolution system, as well as people's bounded rationality and cognitive makeup, deter them from changing their knowledge. Thus, the magnitude of knowledge change also indicates decision makers' flexibility and willingness to change the status quo. Carlile (2002) points out that when decision makers find knowledge that proves useful, they tend to continue to rely on it and are less willing to update it to accommodate new developments. For example, although executives at GE Lighting might receive quarterly syndicated data on retail lighting sales and preferences, not all of them will examine all or even some of the data all the time, and they are likely to interpret the data differently. The executives' degree of attention to the data and their interpretation of it will be reflected in their knowledge about the market and its change over time.

Furthermore, decision makers may or may not update their knowledge on the basis of communication and information exchange with others. Thus, at any given point, they will have a different extent of shared knowledge about consumers retail lighting preferences. This research adopts the theoretical perspective of Hoopes and Postrel (1999, p. 838), who define shared knowledge as "facts, concepts, and propositions [that] are understood simultaneously by multiple agents. The word 'shared' is used as an adjective rather than a participle; that is, we mean only that the knowledge held by two or more individuals is the same, not necessarily that one person has communicated it to the other.... Common knowledge might be a more precise expression, but that term already has a specific technical definition in game theory, as well as a different colloquial meaning." Thus, shared market knowledge among a group of decision makers' is the extent of overlap in individual decision makers market knowledge.

In terms of studying innovation, research in marketing and management has focused primarily on innovation output, as measured by new product success (e.g., Ayers, Dahlstrom, and Skinner 1997) and time to market (Ittner and Larcker 1997). Research in economics has focused more on innovation input, such as research and development (R&D) investments (Miyagiwa and Ohno 1999) and innovation effort (Cheng and Tao 1999; Cohen and Klepper 1996a, b). Because product innovations arise only if a particular level of effort is exerted to bring them to market, the generative process by which innovations emerge needs to be considered. Thus, I focus on innovation effort in this investigation.

The Effect of Market Knowledge on Innovation Effort

Both researchers and practitioners recognize in broad terms that knowledge is an important factor in the creation of competitive success over time (Carlile 2002). For example, researchers studying organizational memory and absorptive capacity (Cohen and Levinthal 1990; Moorman and Miner 1997) have argued, albeit at a macro level, that accumulation of strategic information increases the capacity of organizations to interpret new information and to make strategic decisions, especially if the new information is related to the previously accumulated information. Research in cognitive psychology has shown that people have a fixed capacity to process information (Sternberg 1996), which means that accumulation of strategic information may diminish the ability to interpret new information, unless the new information is related to what has been previously learned. These perspectives together suggest that existing strategic information is more likely to facilitate the interpretation of new information if the existing information is processed, internalized, and converted to knowledge. In other words, market knowledge (rather than information) has positive effects on innovation effort.

H[sub1]: Market knowledge has a positive effect on innovation effort.

The Effect of Change in Market Knowledge on Innovation Effort

In discussing the development of new products, Webster (1997, p. 52) argues that "the product is a variable tailored to changing needs of carefully selected customers. As the customer changes, so must the product, and the organization that provides that product must have built-in flexibility and adaptiveness." Researchers who study the evolution of markets suggest that competitive advantage ensues from a focus on change rather than on static market conditions (e.g., Dickson 1996; Senge 1993). Similarly, Teece, Pisano, and Shuen (1997) argue that organizations should focus not on their current positions and market capabilities but on dynamically changing these over time. Furthermore, it has been strongly argued that the only way to achieve a sustained competitive advantage is through continuous innovation (Schumpeter 1942). These streams of work suggest that the enhancement of the organizational innovation effort requires constant monitoring of the changes in market conditions. Through the monitoring of market condition changes, decision makers enhance their market knowledge. Thus, the degree to which strategic decision makers' knowledge about key marketplace factors changes over time influences organizational innovation effort. For example, if decision makers believe that customer preferences have changed substantially, their innovation effort at that point in time will be more extensive than when they perceive customer preferences as having changed to a lesser degree. A change in knowledge about the actions of close industry competitors has a similar effect: If decision makers perceive movement on the part of their competitors, they are more likely to expend effort innovating than they would in the absence of competitors' actions.

However, how much change in market knowledge stimulates innovation is likely to depend on the original level of market knowledge. Consider two firms that evidence identical changes in market knowledge. If these firms begin at different levels of knowledge, they will have different returns from their knowledge change, and the advantage will be with the firm that has the higher original level of market knowledge, because a firm's original knowledge serves as a basis for integrating new knowledge. Teece (2001, p. 129) argues that "Knowledgeable people and organizations can frame problems and select, integrate, and augment information to create understanding and answers --[that is,] the interpretation of information and its consequent use are determined by the existing knowledge base." A firm that has little knowledge but demonstrates a change in knowledge does not possess a foundation to use this newly acquired knowledge. Therefore:

H[sub2]: A change in decision makers' market knowledge (a) results in an increase in innovation effort and (b) moderates the impact of the level of market knowledge on innovation effort.

Direct and Moderating Effects of Shared Market Knowledge

Prior conceptual work has suggested that the creation and sharing of new knowledge is essential to firms' fostering innovation (Chan and Mauborgne 1997) and building competitive advantage (Senge 1993, 1997). Hoopes and Postrel (1999) argue that unique patterns of shared knowledge are an important source of competitive advantage because they are difficult to purchase and take time to develop. Likewise, in a discussion of learning among strategic decision makers, Argyris and Schön (1978, p. 20) suggest that "[strategic decision makers ] work as learning agents is unfinished until the results of their inquiry ... are recorded in the media of organizational memory." Essentially, this will not occur until shared knowledge is in evidence. Finally, Senge (1993, p. 186) comments that "the most crucial mental models in any organization are those shared by key decision makers."

Despite the widespread belief that shared knowledge enhances innovativeness, empirical evidence is sparse. In one of the few studies that examine actual shared knowledge, Hoopes and Postrel (1999) find that shared knowledge reduces glitches in new product development. Notably, they find that shared knowledge helps significantly reduce project delays and costs associated with lost customer goodwill. Miller, Burke, and Glick (1998) also imply that shared market knowledge offers potential benefits for innovation. Although they do not explicitly focus on actual shared knowledge, they find that cognitive diversity in decision-making teams inhibits strategic long-term planning. Overall, however, researchers have not investigated effects of actual shared market knowledge on important organizational phenomena such as innovation and performance. Instead, they have referred to key respondents' perceptions of consensus about strategic issues as shared knowledge and to diversity in team composition (in terms of demographic factors and functional background) as heterogeneity in organizational knowledge.

I further propose that a high degree of shared market knowledge among decision makers intensifies the effect of the knowledge on the exerted innovation effort. Although decision makers' unified knowledge of market conditions may benefit the application of that knowledge in fostering innovation, wide acceptance of inaccurate information may hinder innovative activities. This reasoning suggests an interaction between the level (accuracy) of decision makers' market knowledge and the extent of shared knowledge among them. Finally, changes in decision makers' knowledge of customers and competitors are likely to stimulate or dampen innovation effort, depending on the extent to which the new knowledge is shared among the decision makers. In other words, changes in market knowledge are likely to intensify the need to innovate, but this relationship will be stronger or weaker depending both on whether market knowledge fosters or subdues innovation effort and on the extent of shared knowledge among decision makers.

H[sub3]: An increase in shared knowledge about customers and competitors has a positive effect on innovation effort.

H[sub4]: The extent of shared knowledge about customers and competitors moderates the impact of (a) market knowledge and (b) changes in market knowledge on innovation effort.

Several forces constitute the market landscape that shapes the coevolution of market knowledge diffusion, firm innovation, and firm performance. To isolate the effects of market knowledge dynamics on innovation effort and the subsequent effect on performance, it is necessary to consider the role of these forces.

Competitive Interdependence

Prior research has shown that a firm's own learning is not the only factor that determines which strategies (e.g., investing in innovation) it adopts and how successful it is; firms also are influenced by the evolving strategic actions and performance of their competitors (Ansoff 1984; Porter 1995). Two firms may implement similar strategies or evidence a similar level of market knowledge, change in knowledge, or knowledge sharing, but they may recover different returns on their strategies simply because they are differently embedded in the competitive landscape (see Dickson 1992). Although the literature shows that the degree of competition and imitation in a firm's market affects its decision to invest in learning, research on organizational and managerial knowledge does not incorporate this competitive aspect (Grandori and Kogut 2002).

Inertia

Against the backdrop of research that has acknowledged the ubiquity of inertia in innovation effort (Bayus, Erikson, and Jacobson 2003; Geroski, Machin, and Van Reenen 1993; Greve 1999, 2002; Miller and Chen 1994), I recognize that firms differ widely in their ability to scan the environment, to make correct cause--effect inferences in environments characterized by the delayed effects of prior strategic actions, and to take action based on these inferences (Senge 1993). Similarly, although prior learning through the adoption and replication of existing action is crucial for future actions (Hargadon and Fanelli 2002), firms differ in their abilities to leverage previous innovation effort. Therefore, I capture each firm's individual proneness to inertia, which reflects the extent to which each firm's current levels of innovation effort depend on its previous levels.

Feedback Effects: The Role of Satisfaction with Past Performance

Drawing from the conceptualization of the cyclical process of market evolution, I recognize that market performance feedback shapes decision makers' satisfaction with past performance, which continuously drives the pattern and speed of their knowledge acquisition. In turn, this influences future strategic actions, such as innovation effort and resultant performance. However, the effect of satisfaction with past performance can be quite complex. On the one hand, prior success (failure) may breed more success (failure) as a result of self-reinforcing mechanisms (see March and Sutton 1997). It is this "nothing succeeds like success" logic that underlies the Schumpeterian hypothesis: In competitive markets, firms whose products enjoy high growth in demand are likely to become more innovative because they are motivated to protect the monopoly power that success engenders (Dosi 1988). On the other hand, success may induce complacency, and failure would result in problem-solving activities aimed at reversing poor performance; these countervailing tendencies then precipitate feedback effects. Support for the latter perspective comes from research in marketing (Chandrashekaran et al. 1999) and organizational behavior (e.g., Audia, Locke, and Smith 2000; Isen and Baron 1991; Miller and Chen 1994). For example, Chandrashekaran and colleagues (1999) find that increased new product diffusion rates negatively affect future innovation generation in a firm. They offer but do not test the perspective that decision makers are prone to success-driven complacency, which arises in part from their satisfaction with the performance of existing products in the marketplace. This perspective implies that managerial knowledge of marketplace feedback results in a continually updated sense of satisfaction with firm performance, which ultimately shapes the extent of innovation effort over time. Overall, the previous two perspectives suggest a U-shaped effect of satisfaction with past performance, such that innovation effort is highest at lower and higher levels of satisfaction.

H[sub5]: Satisfaction with performance has a U-shaped effect on innovation effort.

Apart from the question of how specific aspects of market knowledge diffusion and their interplay at the level of individual decision maker influence innovation effort over time, an important issue is the following: When does innovation effort pay off? I subsequently discuss the mechanism that underlies firm performance, specify a model that tests the translation of innovation effort into performance outcomes, and propose that total shared market knowledge moderates the impact of innovation effort on performance.

What Underlies Dynamic Firm Performance?

A firm may have the market knowledge and resources to achieve a certain level of performance, but it may fail to reach its potential for various reasons, including weak deployment of resources, poor implementation, unexpected competitive actions, and stochastic factors such as random opportunities and obstacles. Furthermore, as I discussed previously, firms differ widely in their ability to make correct cause--effect inferences in environments characterized by the delayed effects of prior strategic actions. Thus, when facing a gap between potential and realized performance, firms differ in their ability to sense such shortcomings and to take action to improve their performance. Although some firms may be acutely aware that their resources should be translating into better performance, previous research indicates that most firms are "gripped by inertia, which dulls and retards responses in a dynamic market" (Chandrashekaran et al. 1999, p. 97). Indeed, the concept of inertia-induced underperformance is the basis of most of the research that attempts to explain why firms that have ample resources in one technological regime often fail to lead in another (e.g., Porter 1995).

Inertia manifests itself in dynamic firm performance by causing realized performance at time to fall continually short of a potential achievable performance at time t. That is, at any point in time, realized performance only partially adjusts to the discrepancy between previously realized performance and potential performance. Moreover, because firms are different, the rate of adjustment is firm specific. Consistent with Leeflang and colleagues (2000), I specify the following autoregressive partial-adjustment model (see the Appendix):

( 1) PERF[subit] = ρ[subi]PERF[subi,t - 1] + X[subit]β - ρ[subi]X[subi,t - 1]β + eta;[subi,t]

where PERF[subit] denotes the performance of firm i at time t, and η[subit] ∼ N(0, σ[subi, sup2]. Four issues warrant comment. First, in this specification, when ρ[subi] = 0 (the adjustment rate is large), there is no inertia. In turn, when ρ[subi] = 1, there is no adjustment in performance, and a condition of complete inertia exists. Second, each firm has a unique volatility associated with performance (note that the variance of η[subit] is firm specific). Third, cov(η[subi], η[subj] = σ[subij, sup2] This accounts for the unique interdependence in performance across firms. Fourth, although I assume a certain specification for η[subit], I explicitly test for serial correlation.

When Does Innovation Effort Pay Off?

Prior research in economics has found that innovation effort has a positive effect on performance (for a review, see Cohen and Klepper 1996a, b). However, a meta-analysis in the marketing literature (Szymanski, Bharadwaj, and Varadarajan 1993) does not find that innovation effort (in the form of R&D) has a positive impact on new product outcomes. I advance a contingency perspective to explain the effect of innovation effort on performance.

Organizational theorists have long acknowledged that knowledge resides in individual decision makers (Simon 1991). However, a person does not think in isolation but interacts with others, so that at the level of the strategic team and organization there is varied alignment of individually held schemas (Hargadon and Fanelli 2002). Reverse alignment of schemas ensures continual knowledge sharing, which enables decision makers to act with unity of purpose. In the marketing literature, scholars have built on these streams of research and have argued that "market knowledge is not fully captured in a usable form until the lessons and insights are transferred beyond those who gain the experience" (Day 1994, p. 44). Guided by this line of research, I suggest that the total shared market knowledge facilitates the translation of innovation effort into performance. A firm that has a high level of total shared market knowledge acts with a greater unity of purpose in making investment decisions than does a firm that has a low level of total shared market knowledge. Thus, I expect that an interaction of innovation effort and total shared market knowledge shapes firm performance. Prior conceptual research has recognized that the sharing of knowledge leads to performance advantages (e.g., Argote 1999; Senge 1993, 1997). However, this research has conceptualized a main effect of shared knowledge on performance and has not tested this empirically. For example, Frank and Fahrbach (1999, p. 254) maintain that processes associated with the pattern of interaction and the act of information sharing among organizational members "represent the geological forces that shape the landscape on which organizational productivity is built." Furthermore, Cannon-Bowers, Salas, and Converse (1993) suggest that shared mental models contribute to more effective teams.

H[sub6]: The impact of innovation effort on performance increases as the extent of total shared market knowledge in a firm increases.

Control Variables

Prior research has recognized the influence of firm size on innovation and competitive processes (e.g., Cohen and Klepper 1996a, b; McKee, Varadarajan, and Pride 1989). Resource-based views of the firm (e.g., Barney 1991), including the resource-advantage theory of competition (Hunt and Morgan 1995), suggest that resources possessed by organizations lead to positions of competitive advantage, which eventually translate into innovation-driven superior performance. Others suggest that small firms are more innovative than large firms because they are less prone to bureaucratic inertia (e.g., Baldridge and Burnham 1975) or that the very success that yields greater financial resources serves to create complacency, which hinders subsequent innovation (e.g., Audia, Locke, and Smith 2000). Consequently, I control for firm size in my model of innovation effort.

Prior work in economics also supports the notion that firm size influences return on innovation (Cohen and Klepper 1996a, b). Therefore, in addition to a main effect of firm size, I expect that there is an interaction between innovation effort and firm size. Inclusion of this interaction allows for an examination of the relative impact of firm size (compared with the impact of shared knowledge) in the translation of innovation effort into firm performance, and it helps empirically uncover sources of competitive advantage that prior research might have ignored. Finally, prior research in marketing has also documented that strategic orientation has an effect on firm performance (e.g., Gatignon and Xuereb 1997; Voss and Voss 2000); thus, I include strategic orientation in the study.

To address the focal research questions, a controlled setting is required that allows for the explicit capture, over time, of market knowledge and its role in actualizing innovation in a dynamic system of market evolution. I thus turned to a quasi field experiment: the Markstrat3 simulation (Larreche and Gatignon 1999), which has been widely used as an empirical setting in prior research on managerial decision making (Clark and Montgomery 1998, 1999; Glazer, Steckel, and Winer 1989, 1990, 1992; Van Bruggen, Smidts, and Wierenga 1998).

The Markstrat setting is highly suitable for this research for several reasons. First, the simulation provides objective information on all aspects of the marketplace. This is crucial for assessing decision makers' market knowledge and thus meaningfully capturing the level of accurate market knowledge, the change in knowledge, and the extent of shared knowledge. In addition, this research requires repeated measures of the various dimensions of each decision maker's knowledge over time, which precludes a cross-sectional survey method.

Second, because the simulation makes it possible to observe the vector of independent variables before performance is known to the firms, there is no problem of contemporaneous correlation between the error term and the vector of independent variables. Although in principle the research task can be accomplished by collecting longitudinal data through repeated interviews in an industrial setting, the issue of contemporaneous correlation remains, because it would be impossible to observe the multitude of relevant variables (e.g., tangible and intangible performance incentives) that simultaneously affect the outcome and the variables of theoretical interest. However, with a simulation, identical incentives are established for all participants, which is an essential requirement for the design of this study.

Third, the simulation can be set up such that all teams begin in identical positions. Consequently, variations that emerge in subsequent strategy and performance can be more easily linked to distinct capabilities across firms. Fourth, the structure of all firms is identical and simple: A team of four to six members is responsible for all decisions. This structure controls for macrostructural effects on decision making and innovation.

Participants are assigned to one of six teams, and each team is responsible for managing a firm in the Markstrat industry over a certain number of time periods (ten in the main study, which corresponds to ten years). At the start of the simulation, all firms have two products in the market. Throughout the course of the simulation, firms must make decisions about production, advertising, sales force, distribution, pricing, and product positioning for each product in each period. Firms can also purchase any or all of a set of market research reports each period. They can introduce new products (after the successful completion of R&D) and withdraw old ones from the market.

Design

I performed two pretests using different industry settings, different subjects, and different points in time to assess the feasibility of data collection in such a setting, to examine how well the models generated estimates that could serve as a basis for subsequent triangulation, and to make necessary improvements in the data collection protocols. Pretest 1 involved 17 part-time and full-time MBA students at a large Midwestern university. The participants had an average of 8.7 years of work experience and an average age of 32. The data collection process was based on nine decision periods and was divided into two phases. In the orientation phase (Periods 1-3), subjects focused on learning the various aspects of the simulation, and in the data collection phase (Periods 4-9), they completed, on an individual basis, questionnaires aimed at capturing their market knowledge at each point in time. The panel data from this pretest came from 17 decision makers, in five teams, observed over six continuous time periods. Pretest 2 helped refine the measures for knowledge on customer preferences and competition at the micro level. The design was similar to that of the first pretest. Participants were 12 full-time MBA students at a different large Midwestern university, with an average of 3.5 years of work experience and an average age of 24.4. The panel data from this pretest came from 12 decision makers, in four teams, observed over nine continuous time periods.

For the main study, six teams, each of which contained four or five MBA students from a large Midwestern university, participated in a Markstrat3 simulation in which identical starting positions were used. All teams competed in only one industry (Sonite) for the entire duration of the simulation. (Entry into a second industry [Vodite] was prevented by making conditions for entry impossible to meet.) Over two-and-a-half months, 25 participants provided data on their knowledge and perceptions of the marketplace. The average age of the participants was 27.2 (standard deviation 3.61, range 21 to 33 years), and the average work experience was 5.7 years (standard deviation 3.67, range 1 to 14 years).( n1) Before the actual simulation started, the participants were familiarized with the software and the mechanics of the simulation through a trial run of three decision periods.

Questionnaires aimed at capturing relevant aspects of each decision maker's knowledge structure were administered every period in Periods 2-9. While participants were completing the questionnaire, they were not allowed to consult their notes or their team members. To alleviate problems of diminishing interest and automatic response patterns, which are typical of repeated-measure questionnaires, participants were continually reminded that there were no right or wrong answers and that the questions aimed to uncover their perceptions of the dynamics in the market. In addition, the content of the questionnaires was varied (but contained the same key measures every decision period) to maintain participants' interest, and the questions captured the diversity of decision-making aspects well. Immediately after each team had made its decisions and before it received performance feedback, each decision maker's knowledge of the product-market factors and perceptions of the previous period's performance were measured. Therefore, the panel data for the analysis came from 25 decision makers, in six teams, observed over eight continuous time periods, which yielded 48 organization-period observations.

I also conducted three validations studies. In the first, decision makers participated in Markstrat without being surveyed at all during the simulation. This served two purposes. First, it validated the underlying generative mechanism of competitive dynamics, and second, it tested whether the process of repeated data collection influenced the dynamic system of competition. The panel data from this study were collected from 28 MBA students (not from the same university as the students who participated in the main study), with average work experience of three years, in five teams over nine time periods.

Because I also aimed to assess the robustness of the results, both by attempting to replicate them and by examining the cross-industry predictive ability of the estimates, I conducted two more validation studies in two different Markstrat industries. The second validation study involved 36 MBA students (average work experience of 7.3 years) in six teams over nine time periods, and questionnaires were administered for seven periods. The third study involved 15 MBA students (average work experience of 3.3 years) in four teams over nine time periods, and data were gathered for four periods.

Measures

Knowledge. Conceptually, knowledge is defined as the extent of the accurate detection of a task's properties; in the context of this study, it involves the accurate detection of customer preferences and competition. Specifically, in the Markstrat industry setting, critical properties are accurate identification of customer segment preferences or ideal points and competition. Because all teams ordered market research reports and spent time analyzing them, I considered the distinction between information and knowledge.

Conceptually, knowledge and information are different but related. In general, knowledge is considered interpreted information that is anchored in people's beliefs and commitment (Huber 1991). Knowledge also comprises such cognitive elements as beliefs, understanding, interpretation, and integration with previous knowledge. In the words of Bertels and Savage (1998, p. 17), "information only comes alive by our interpretation.... [W]e create meaning by distinguishing and valuing information." Although decision makers obtain market reports that list the ideal points and spend time reading reports, information needs to be integrated into their schemas and into their prior understanding of the market for it to become knowledge. If this information is not integrated with a person's existing understanding of the market and remembered for later retrieval, it is also less likely to be subsequently used. Moreover, teams (and decision makers) differ in terms of their ability to process, understand, and interpret this information. Yet the development and execution of successful strategy in Markstrat (also in general) requires the integration of many pieces of information. For example, information that addresses different elements of the competitive environment is located in different reports. Thus, decision makers need to be able to process and remember a piece of information to relate it to another when they encounter it and need to integrate it. In addition, successful performance requires marketing products that satisfy segment preferences. The greater the intensity of competition in a product market, the greater is the impact of the discrepancy between perceived characteristics of products and the "ideal" points of the target segment on sales. If decision makers understand the importance of this aspect of product marketing, target-segment ideal points will be highly salient to them, actively used, and more easily remembered.

Furthermore, teams also differ in diagnosing both the relative importance of different information that is available in the reports and the importance of the degree of accuracy of customer knowledge necessary to compete (and/or deemed important for their strategy). Given the availability of information, teams may have relatively inaccurate knowledge of the ideal points. For example, this occurs under the following three conditions: First, this occurs when teams believe that delivering a product that corresponds exactly to the ideal points of the segments is not needed and that being "close enough" is sufficient to compete in accordance with their objectives at a particular point in time (there is a variation in what "close enough" is; in the extreme case, it can mean dichotomization of the market in terms of high-and low-end products), and variation can exist across and within teams across time. Second, when teams believe that performance in the previous period suffered because they overlooked competition, they may spend more of their time analyzing competition and less time studying changes in customer preferences (decision makers also differ in their abilities to diagnose cause--effect relationships). Given fixed capacity to process information (Sternberg 1996) and limited time, their customer knowledge will likely be compromised and less accurate than knowledge of competition, especially if significant changes occurred in the market. Third, inaccurate knowledge can result from teams becoming complacent with their performance (Audia, Locke, and Smith 2000), in which case they are likely to be less sensitive and to pay less attention to changes that take place in the market. In this case, their customer knowledge will reflect previous ideal points, which are not updated and thus are not accurate.

Given the previous considerations, each decision maker's knowledge about customer preferences was assessed through a series of questions that they answered immediately after they had made decisions but before they had seen the performance outcomes. They marked the location of various customer segments' ideal points on a perceptual map, and they indicated the relative importance of each customer segment for their firm strategy and operations by dividing 100 points among the existing segments. The accuracy of their knowledge was judged by the Euclidean distance between the points they specified on the map and the objective ideal points for each customer segment (provided by the simulation), weighted by segment importance as reported by each decision maker. I incorporated segment importance to give more weight to accuracy (or inaccuracy) in knowledge for segments that are more important to the firm's operations. Thus, I measured customer knowledge as follows:

( 2) [Multiple line equation(s) cannot be represented in ASCII text]

where k (= 1, ..., 5) denotes segment; P[subatt1,t] and P[subatt2,t] denote decision makers' perceptions of the ideal values of the two most important product attributes for each segment at time t; IP[sub1,t] and IP[sub2,t] denote the objective ideal values of the two most important product attributes for each segment at time t; and S[subk,t] is the importance of each segment k as indicated by the decision maker at time t.

I assessed knowledge about competition by asking each decision maker at each point in time t (each decision period) to list the names of all competing products in each segment and to identify the most influential competitor in each segment. I employed two measures of knowledge about competition. First, consistent with prior research on competitive monitoring (Clark and Montgomery 1999), I focused on the number of competitors identified and defined competitor knowledge as COMPK[subt] = (number of competitors identified at time t)/(total number of actual competitors at time t).( n2) Second, to account for strength or intensity of competition, I weighted each identified competing product by its distance in perceptual space from the ideal point of the segment designated by the decision maker at time t and by its importance, S[subk,t]. For each decision maker,

( 3) [Multiple line equation(s) cannot be represented in ASCII text]

where k (= 1, ..., 5) denotes segment; j (= 1, ..., m) denotes product; D is a dummy variable that equals 1 if the decision maker has identified product j as a competitor in segment k at time t and equals 0 otherwise; IP[sub1,t] and IP[sub2,t] denote the objective segment ideal values of the two most important product attributes for each segment at time t; CP[subatt1,t] and CP[subatt2,t] denote the product's position in perceptual space on the basis of the two most important product attributes at time t; and S[subkt] is the importance of each segment k as indicated by the decision maker at time t. Analyses with both measures produced the same pattern of results. Finally, I computed the average value of CUSTK and COMPK (COMPK2) across decision makers in each firm to represent, respectively, level of knowledge of customers and level of knowledge of competitors in each firm.( n3)

To validate these measures further, I estimated a panel model with market knowledge (at the level of the individual decision maker) as the dependent variable, individual as fixed (random) effect, time as fixed (random) effect (or accounted for as an autocorrelation effect, ρ = .36), and variables expected to affect market knowledge as independent variables. The results indicate that all variables are significant in the hypothesized direction and jointly explain 29% of the variance, whereas the individual effects explain 18% of the variance. In addition, within the same analysis framework, I tested the effects of involvement (measured as the percentage of effort or time respondents spent on analyzing information and preparing for decisions on their own) and found that it did not have a significant effect on market knowledge (parameter = .014, p < .78). I also examined whether differences in knowledge within a team and across team members were a result of task assignment. The average delegation of tasks within teams was 11.06% of all tasks, which accounts for 1.8% of the variance compared with between-group effects, which account for 10% of the variance. This suggests that differences in market knowledge across team members are not just a consequence of the assigned tasks within teams, nor does the amount of delegated effort fully explain the between team effects..

Change in knowledge. Given the assessment of decision makers knowledge every period, I measured individual change in knowledge about customer preferences (CUSTC) and about competition (COMPC) as |∂CUSTK[subt]/∂t| and |∂COMP[subt]/∂t|, respectively, and accounted for the rate of change in the different market segments. In the analysis, the average value of CUSTC and COMPC across strategic decision makers in each firm represented change in customer knowledge and change in competitor knowledge, respectively.

Shared knowledge. Consistent with standard measures of diversity across respondents, I measured the extent of shared knowledge about customer preferences (CUSTS) as the variation in CUSTK across decision makers in each firm. Specifically, I computed CUSTS as 1/σ[subCUSTK,t], where σ[subCUSTK,t] denotes the standard deviation in the level of customer knowledge across decision makers at time t. I measured the extent of shared knowledge about competition (COMPS) in two ways, both of which were consistent with the two measures of competitor knowledge. First, I measured shared competitor knowledge as the extent of variation in COMPK (number of competitors identified), computed as 1/σ[subCOMPK,t], where σ[subCOMPK,t] denotes the standard deviation in knowledge level of competitors across decision makers at time t. Second, I measured shared competitor knowledge per Equation 3 by including only competing products j that all decision makers in the team identified as competitors in a particular market segment.

Strategic orientation. In line with prior research (Gatignon and Xuereb 1997), each decision maker was asked to allocate 100 points to the following three categories to reflect the categories' relative importance for each time period's investment decisions: competition, consumer preferences, and nature of products. I measured the strategic orientation of the firm at time t as the average difference in the importance of the customer and competition (Day and Nedungadi 1994). Thus, positive values for this measure indicate greater emphasis on customers than on competitors, whereas negative values reveal a greater emphasis on competitors than on customers.

Total shared market knowledge. Following guidelines in prior research (Day and Nedungadi 1994), I measured total shared market knowledge as 1/σ[subORIENT,t], where σ[subORIENT,t] is the standard deviation of the firm's strategic orientation at time t, such that small values of σ[subORIENT] indicate a greater extent of total shared market knowledge. An alternative measure could be based on a combination of the measures of customer and competitor knowledge. The advantages of the measure I used are that it does not assume that there is a particular rule for combining customer and competitor knowledge, and it is an index of overall shared market knowledge, including but not limited to knowledge about customers and competitors. According to prior research (Day and Nedungadi 1994), the extent of overlap in decision makers' mental representation of their firm's strategic orientation reflects the extent of the decision makers shared market knowledge.

Satisfaction with past performance. I assessed individual decision makers' satisfaction with past performance (SAT) after each period's decisions using a seven-point Likert scale, anchored by "strongly agree" and "strongly disagree." The statements that participants rated were "I was satisfied with my team's performance" and "My team did not do as well as I thought it would." The two measures were highly correlated (.734, p < .0001); thus, I averaged scores across decision makers in a firm to represent satisfaction with past performance in that firm.

Firm size. There are many ways to measure firm size, including by number of employees, financial resources, net sales, or total expenditures (e.g., Chandy and Tellis 2000). Given the nature of this study's quasi-experimental setting and processes of interest, number of employees and net sales are not appropriate measures (see Clark and Montgomery 1999). Thus, consistent with the resource-advantage perspective, I measured firm size (SIZE) as the total financial resources (in dollars) available to each firm at each point in time t. (No loans were permitted in the simulation.) In the model of performance, following the lead of Clark and Montgomery (1999), I used total marketing expenditures as an alternative measure of firm size. The advantage of this measure is that it also controls for marketing-mix expenditures, which have been shown to drive performance. However, the pattern of results remained the same across the two measures of firm size. Therefore, I report results from the first measure.

Innovation effort. I measured innovation effort (INNOV) in two ways: by the dollar amount of each team's investment in product R&D and by the ratio of dollar investment in product R&D to total marketing expenditures (including advertising and sales force activity). The first measure captures the level of innovation effort, and the second captures its intensity. Because the two measures were highly correlated (r = .92, p < .0001) and analyses with both produced the same pattern of results, I report results from the first method.

Performance. I measured performance (PERF) in two ways: by the firm's market share, in units sold, at each point in time and by the firm's stock price index. In the context of Markstrat, the firm's stock price index is the most comprehensive measure of firm performance (Larreche and Gatignon 1999). In the context of this study, as with prior studies that used Markstrat, a significant correlation exists between unit market share and net marketing contribution (r = .82, p < .001). Descriptive statistics for all variables in the analysis appear in Table 1.

Analysis Framework

I specify the following framework for estimating the impact of market knowledge diffusion (i and t denote firm and time, respectively):

( 4) INNOV[subit] = X[subit]β + ε[subit],

( 5) ε[subit] = ρ[subi]ε[subit] - 1 + η[subit]

( 6) ρ[subi] = ρ + λ[subi], and

( 7) cov(ε[subit], ε[subjt] = σ[subij]

where X captures the independent variables that are hypothesized to affect the innovation effort (see H[sub1]-H[sub5]),( n4) η is an error term, and var(ε[subit]) = σ[subit, sup2]. Furthermore, ρ[subi] captures the extent of unique inertial tendency in innovation for firm i. Large variation in ρ[subi] (i.e., in λ[subi]) can be treated as evidence that firms are heterogeneous in their inertial tendency in innovation effort. In turn, σ[subij] in Equation 7 captures the interdependence of the innovation effort of firm i and firm j. (Note that σ[subij]/√[σ[subit, sup2] x σ[subjt, sup2] is bounded by 1 and +1 and is a correlation coefficient.) Notably, Equations 4-7 imply the following structure for innovation effort:

( 8) INNOV[subit] = X[subit]β; + ρ[subi](INNOV[subi,t] - X[subi, t - 1]β) + η[subit].

Thus, after accounting for innovation interdependence across competitors, each firm's innovation effort at every point in time is influenced by current levels of the X variables and by the extent of unique inertial tendency in its innovation effort, beyond that explained by previous levels of X.

Depending on restrictions placed on the model parameters in Equations 4-7, models that correspond to different perspectives can be obtained; these are summarized in Table 2. To estimate the models, I employed a feasible generalized least squares procedure (Greene 2002).( n5) I began with the simplest model, M00, and proceeded systematically to free parameters until I obtained the hypothesized Model M[sub12]. To assess model fit, I performed model comparison tests at each stage of the estimation. A comparison of Model M[sub11] with M[sub12] (and M[sub01] with M[sub02]), which I accomplished using χ² [sub(i degrees of freedom [d.f.])], tests for the presence of unique inertial tendency in innovation effort. A comparison of Model M[sub02] with M[sub12] (and M[sub01] with M[sub11]), which I accomplished using χ²[sub[i(i - 1)/2 d.f.]], tests for the presence of interdependence in innovation effort across competitors. I used the same approach to estimate the model of firm performance presented in Equation 1.( n6)

Market Knowledge Diffusion and Innovation Effort

Competitive interdependence and inertia Model M[sub01] (see Table 2) is not superior to model M[sub00] (ρ = .236, χ[sub1, sup2] = .41, not significant [n.s.]). Furthermore, Model M10 is superior to model M[sub00] (χ[sub15, sup2] = 54.93, p < .0001). Finally, a comparison of Models M[sub10] and M[sub12] reveals that the best-fitting model is M[sub12] (χ[sub5, sup2] = 9.80, p < .05; values of ρ across firms were .364, .047, .307, .990, -.192, and -.006). This indicates that firms exhibit a unique inertial tendency in innovation effort and that in accordance with the cyclical model of market evolution, there is significant heterogeneity in the interdependence of innovation effort across firms. Further analyses are based on Model M[sub12].

Table 3 shows all the parameter estimates of Model M12 that I used to test specific hypotheses regarding the effects of customer and competitor knowledge, change in knowledge, shared knowledge, and satisfaction on innovation effort. Because of the significant interactions in the model, I computed the net effect of the key variables at different levels of the other independent variables, and I report the results in Table 4; in each case, I used the Wald test to test significance (Greene 2002).

Effects of customer and competitor knowledge. Table 3 shows a significant interaction among shared knowledge, change in knowledge, and knowledge level. The net effect of each aspect of knowledge diffusion, which provides support for [Hsub1-H[sub4], is shown in Table 4.

Effects of shared knowledge. Overall, Panel A of Table 4 reveals that across knowledge levels and magnitude of knowledge change, increases in shared knowledge about customers and competitors enhance innovation effort, which provides strong support for H[sub3]. For example, for low customer knowledge and high knowledge change, the greater the shared knowledge, the greater is the innovation effort (net effect = 2.936, p < .001).( n7)

Effects of change in knowledge. For customer knowledge (Panel B), increases in knowledge change enhance innovation effort when shared knowledge is high. At low levels of shared knowledge, a change in knowledge has no effect on innovation effort across different knowledge levels. For competitor knowledge (Panel B), increases in knowledge change enhance innovation effort across different levels of knowledge and shared knowledge. However, these effects are weaker for low levels of shared knowledge. For example, at low levels of competitor knowledge and high levels of shared knowledge, the net effect of knowledge change is 1.131 (p < .05).( n8) In contrast, when shared knowledge is also low, the net effect of knowledge change is ten times smaller (.105; p < .1). Thus, I obtain mixed support for H[sub2a] and strong support for H[sub4b], which indicates that shared customer and competitor knowledge moderates the impact of knowledge change on innovation.

Effects of knowledge level. Panel C of Table 4 reveals that increases in the level of customer knowledge enhance innovation effort only when there are high levels of shared knowledge about customers.( n9) The effect of customer knowledge is not significantly (only directionally) enhanced by the magnitude of knowledge change, which does not support H[sub2b]. If there is little shared knowledge in the firm, an increase in customer knowledge has no impact on innovation effort. I obtained the identical result for competitor knowledge: The higher the level of competitor knowledge in the firm, the greater is innovation effort, if and only if shared knowledge about competitors in the decisionmaking team is high. If there is little shared knowledge, an increase in competitor knowledge has no impact on innovation effort. In summary, I obtained strong support for H[sub4a]: An increase in shared knowledge about customers and competitors moderates the impact of market knowledge on innovation effort.

Feedback effects: satisfaction with past performance. I obtained significant main (b[sub9] = 11.50, p < .0001) and quadratic (b10 = .205, p < .0001) effects of satisfaction in support of H[sub5] (see Table 3). Notably, because of the structure of the dynamic model, the effect of satisfaction is beyond any effects of past performance. The overall effect of satisfaction (computed as b[sub9]SAT + b[sub10]SAT²) is negative across all levels of satisfaction, which supports the hypothesis that satisfaction with performance is a complacency-producing mechanism that dampens innovation effort. The results of computation of the net effect of satisfaction on innovation effort (i.e., ∂INNOV/∂SAT = b[sub9] + 2b[sub10]SAT) reveal that at low levels of satisfaction (tenth percentile), the net effect is negative and significant (-4.676, p < .001). Thus, if decision makers are dissatisfied with their performance, an increase in their satisfaction hinders innovation effort. However, as satisfaction approaches its mean value, increases in satisfaction result in an increase in innovation effort (net effect = 1.445, p < .001). I obtained the same effect at high levels of satisfaction as well (net effect = 7.976, p < .0001). Although at the mean and high levels of satisfaction, a unit increase in satisfaction leads to an increase in innovation effort, the overall impact on innovation effort is still negative. Thus, the quadratic effect of satisfaction indicates that the perceived feedback from the market is most damaging to the innovation effort of firms that exhibit moderate levels of satisfaction. Firms that exhibit high or low levels of satisfaction witness a smaller decrease in innovation effort.

From Innovation Effort to Performance: The Role of Total Shared Market Knowledge

To ensure that the errors in Equation 1 were not autocorrelated, I extracted the residuals after estimating Equation 1 and tested them for the presence of autocorrelation. Results indicated that the errors were largely white noise (ρ = .35, χ[sub1d.f., sup2] = 1.11, n.s.). An examination of model fit revealed a significant inertial tendency in performance (ρ = .63, p < .0001) but no significant difference in the inertial tendency across firms (χ[sub5, sup2] = 3.496, n.s.), that is, no significant difference in ρ[subi] and the adjustment rates across firms. There was also support for heterogeneity in performance interdependence across firms (χ[sub15, sup2] = 39.12, p < .0001). Overall, the model explains 80% of the variation in performance.

Drivers of performance. Table 5 contains all the parameter estimates of the retained model. Observe in Table 5 that a similar pattern of effects emerges across measures of performance. For the purposes of concision, the following discussion focuses only on the results derived from the market share analysis. The results indicate that there is no significant main effect of innovation effort on performance (b[sub1] = .090, n.s.).)( n10) However, consistent with H[sub6], there is a significant interaction between innovation effort and total shared market knowledge (b[sub2] = 5.438, p < .0001) and between innovation effort and firm size (b[sub3] = .231, p < .01). Furthermore, there are significant main effects of firm size (b4 = .921, p < .0001) and strategic orientation (b[sub6] = .121, p < .0001), which corroborate prior findings in the literature.( n11) Last, it appears that total shared market knowledge per se has no direct effect on performance (b5 = .590, n.s.). On the basis of the results, the return on innovation (i.e., the net effect of innovation effort on performance) can be expressed as follows: Return on innovation = ∂PERF/∂INNOV = .090 + 5.438 x shared market knowledge -.231 x firm size.

There is strong support for H6, which indicates that the total shared market knowledge helps translate innovation effort into performance. Innovation effort results in positive returns only if it is accompanied by a high level of shared market knowledge among decision makers. Innovation effort results in negative rather than positive returns for large firms if total shared market knowledge is low (at the ninetieth and tenth percentiles for firm size and shared knowledge, respectively; ∂PERF/∂INNOV = -.026, p < .001). Notably, small firms do not witness such a negative effect: Innovation effort has no influence on performance when the level of total shared market knowledge is low; for example, at the tenth percentile for firm size and shared knowledge, ∂PERF/∂INNOV = -.003, n.s. Small firms evidence positive returns on their innovation effort when shared market knowledge reaches the fifty-fifth percentile of its distribution (at the tenth and fifty-fifth percentile for firm size and shared knowledge, respectively, ∂PERF/∂INNOV = .08, p < .06), whereas large firms show positive returns only after shared market knowledge is greater than the ninetieth percentile (∂PERF/∂INNOV = .012, p <.06). Furthermore, the results suggest that though large firms enjoy higher performance, this effect becomes weaker as innovation effort increases (observe the negative interaction between firm size and shared market knowledge). This finding implies that larger firms are less efficient in obtaining a return on resources used for innovation. Thus, although greater firm size appears to be an advantage in the generation of positive performance returns, it is also an impediment to efficient innovation effort.

Validation and assessment of theory. I conducted several validation exercises with data from three validation studies to assess the impact of repeated data collection on the generative dynamic system of competition, to attempt to replicate the substantive findings from the performance model, and to test the cross-industry predictive ability of the performance model. I then examined the nature of substantive findings that might have emerged had I ignored heterogeneity in competitor interdependence in the area of innovation effort and firm performance.

The impact of repeated data collection on the generative mechanism of competition. The focus was on validating the two aspects of dynamic competitive markets: partial adjustment of performance over time and the interdependence in competitors performance. Consistent with the validation objectives, I focused on models of firm performance without any covariates. Table 6 reports the key results. Observe in Table 6 that across three different industries at different times and involving different levels of data collection, support for the underlying generative mechanism emerges. There is partial adjustment of performance (average and standard deviation of ρs are similar across the three industries), and competitors show significant interdependence in performance. In Validation Study 1, there were no periodic intrusions on the simulation to collect data, yet the same baseline process is evident. Thus, the collection of periodic data in the main study appears not to have influenced the generative mechanism investigated.

Replication of substantive findings. Table 5 presents results of the replication of the original study. The hypothesized variables show a pattern of effects on performance that is similar to the pattern in the original study, thus replicating the previously reported findings.

Cross-industry predictions. I used the estimates of β and ρ from the original study to predict firm performance for the industries in the second and third validation studies. I was interested in determining whether I could use the obtained insights to predict performance in a different industry. The results of the cross-industry predictions are summarized in Table 7, along with the same-industry predictions. I obtained correlations of .64 and .92 between actual and predicted performances in the second and third validation studies, respectively. The results provide further evidence of the validity of the hypothesized variables impact on performance.

Impact of ignoring competitor interdependence. In terms of innovation effort, failure to account explicitly for interdependence across competitors produces a null effect for interactions that involve customer knowledge, change in customer knowledge, and shared customer knowledge, as well as for interactions that involve competitor knowledge, change in competitor knowledge, and shared competitor knowledge. In terms of firm performance, failure to account explicitly for interdependence across competitors produces a null effect for the interaction between innovation effort and total shared market knowledge as well as for the interaction between innovation effort and firm size.

This article offers a conceptualization and a longitudinal empirical assessment of the dynamic process that governs the translation of knowledge about customers and competitors into innovation effort and performance. First, I mapped the mechanism by which three aspects of knowledge (knowledge level, knowledge change, and extent of shared knowledge about customers and competitors) influence innovation effort. Second, I uncovered the role of total shared market knowledge and firm size in the translation of innovation effort to performance. The proposed model explicitly incorporates critical aspects of the dynamic competitive process, including firm-specific inertial tendency in innovation, heterogeneity in interdependence of innovation across competitors, feedback effects reflected in satisfaction with past performance, and partial adjustment of firm performance over time. Results from a longitudinal quasi field experiment and series of validation studies provide strong support for the theory.

Limitations

First, although prior studies that have used Markstrat as a research platform have typically collected data at the team level, most of the measures in this study required data from all the team members and over time. The use of frequent collection periods placed constraints on the number of participants involved in the study and on the number of variables involved. The benefits of conducting longitudinal quasi field experiments were balanced by smaller samples that provided a relatively conservative test of the theory.

Second, following Clark and Montgomery (1999), I measured firm size in terms of available financial resources rather than number of employees or net sales (see Chandy and Tellis 2000). Although I replicated the study's results using marketing expenditures as a measure of firm size, the existence of multiple methods to measure firm size warrants attention in further research. The use of other measures in different contexts might further clarify the role of firm size in innovation and performance. In addition, this study could have benefited from an investigation of additional aspects of market knowledge (e.g., it might have explored knowledge about the features of competing products), but this would have added significantly to the already substantial data collection task.

Third, I tested the theory in the context of simulated industries. As a quasi-experimental setting, Markstrat does not provide the same environment as an actual company. The simulated environment made it possible to control for organizational structural factors that might influence the studied relationship, but this required the assumption that the strategic decision-making team drives the strategic course of the organization. Thus, the simulation does not address the diffusion of market knowledge throughout the organization. However, despite the limitations inherent in the use of a simulation, the following aspects enhanced the validity of the study: ( 1) replication of the baseline generative system of competition at different points in time over two-and-a-half years, involving different levels of data collection in the context of four different industries; ( 2) replication of the study's substantive findings on firm performance; ( 3) use of an efficient feasible generalized least squares estimation, which made it possible to obtain consistent estimates of the parameters of interest (Greene 2002); ( 4) generation of solid cross-industry predictions of firm performance; ( 5) all the variables included in the analysis having a significant role in the phenomena explored, either as main effects or in interactions involving other variables, though sample size influenced the power of the estimation; and ( 6) examination of how the collection of periodic individual data might affect the behavior of decision makers, which revealed no evidence that repeated data collection adversely affects the findings. Replication of this study in nonsimulated industries would enhance the generalizability of the findings.

Fourth, the study did not directly measure decision makers expectations about their decisions; instead, it measured satisfaction with past performance. According to the expectation-disconfirmation theory of satisfaction (e.g., Oliver 1980), satisfaction is expected to mediate the effect of prior expectations on strategic decision making. According to this perspective, satisfaction with past performance should contain information about expectations associated with the decisions as well as the comparison of the expectations with performance.

Fifth, although for maximum benefit shared knowledge should comprise knowledge not only about the market but also about the company, products, financial issues, objectives, strategies, internal capabilities, processes, and so on, this article focuses strictly on shared market knowledge. It does not address other aspects of shared knowledge in general or differences in perspectives based on functional responsibility.

Innovation Effort

This study reveals that three aspects of market knowledge diffusion (market knowledge, change in market knowledge, and shared market knowledge) and their interplay over time shape the extent of innovation effort. The results indicate that mere possession of accurate knowledge about customers and competition does not lead to enhanced innovation. Instead, change in market knowledge and shared knowledge assume key roles in transforming market knowledge into innovation. The results also support prior research conjectures (e.g., Chandrashekaran et al. 1999) that satisfaction with past performance is an inertia-producing mechanism that suppresses innovation over time. The quadratic effect of satisfaction, which shows that firms that exhibit high or low levels of satisfaction demonstrate a smaller decrease in their innovation effort, is perhaps indicative of successful firms' motivation to protect their market power and of poorly performing firms' desire to change their fortune. Firms characterized by moderate levels of satisfaction are perhaps plagued by the desire merely to stay afloat; they may avoid changing their behavior because they may be more uncertain about the drivers of market success. Thus, moderate levels of satisfaction may be a manifestation of uncertainty about the market, and this uncertainty may result in failure to innovate. This constitutes a worthwhile avenue for further research.

The results also suggest that innovation effort takes shape over time under the influence of two opposing forces: market knowledge diffusion, which propels innovation, and satisfaction with past performance, which hinders it. Moreover, in the current study, the opposing forces assumed equal importance in the innovation-generation process, explaining approximately equal amounts of variation in innovation effort. Furthermore, the results offer evidence of unique inertial tendency in innovation effort and of interdependence in innovation effort across competitors. Failure to control explicitly for such strategic interdependence produces an inferior fit to the data and null effects for the impact of many variables. This points to the need for further research to consider these effects when investigating the antecedents of innovation across firms. It also raises worthwhile questions for further research. For example, how do firms learn about the innovation effort of their competitors, and what are the inference processes at work? What are the factors that determine the extent of inertial tendency in innovation effort and heterogeneity in inertia across firms? Some factors that further research might consider include individual-difference variables, such as decision makers proneness to inertia or risk taking, and aspects of the decision-making process, such as strategic conjecturing about competitors and customers, beliefs about market dynamism, and effectiveness of implementation. Bayus, Erickson, and Jacobson (2003) initiate such an inquiry by examining the number of new products introduced as a potential determinant of inertial patterns in profits across firms in the personal computer industry, but their study finds no effects.

Performance

With regard to the relationship among shared market knowledge, innovation, and performance, the results of this study suggest that shared market knowledge helps smaller firms actualize better returns for their innovation effort, which means that shared market knowledge may be a source of competitive advantage for resource-strapped firms. Specifically, large firms, which experience inertial effects of their financial resources, need to achieve higher levels of shared market knowledge to enjoy the same positive returns as small firms. Strikingly, for small firms in this study at the mean level of shared market knowledge, an increase in innovation effort by $1 million results in a return of 1.3% market share points. Large firms experienced almost the same return (1.2% market share points) only at the ninetieth percentile of shared market knowledge. Studies across different contexts should be conducted to replicate the results and to identify limiting conditions based on industry factors and dynamics. Overall, this model shows that the impact of innovation effort on performance depends on shared market knowledge and firm size. Still unknown are the mechanisms that firms can employ to facilitate the efficient and effective sharing of market knowledge. Given the cost of attaining shared market knowledge, what are the structural and cognitive impediments to achieving optimal shared knowledge? Are the patterns of inertia in innovation and firm performance also prevalent in the processes by which individual decision makers learn and share knowledge with one another? These are questions worthy of future research endeavors.

Conclusion

In a significant departure from prior work, this research adopts a dynamic perspective of the simultaneous evolution of market knowledge and innovation. It also contributes to the debate about the role of market knowledge accuracy in strategic decision making. Although I used a quasiexperimental setting to test the conceptual model (which requires replications in different industrial settings), the research clarifies the role of market knowledge and suggests that market knowledge, by itself, has no positive effects on innovation effort or performance. When market knowledge is updated (i.e., changed on the basis of individual interpretation of the market) and shared among decision makers (i.e., there is common understanding of the relevant market knowledge in the team), it leads to increased innovation effort and higher returns to innovation. Furthermore, the results reveal that though innovation effort is sensitive to shared knowledge about both customers and competitors, it is more sensitive to competitor knowledge than to customer knowledge. Although the spectrum of this response no doubt depends on the industry setting, assessment and management of the returns to market knowledge should take into account the dynamic impact of knowledge change and sharing in the strategic decision-making team over time. The results also imply that this might be more important for resource-constrained than for financially endowed firms.

Finally, this research has built on the momentum created by prior research in the marketing literature on organizational learning, strategic orientation, and organizational memory. I hope that it generates further examination of the dynamic processes that underlie market knowledge diffusion and the nature of its consequences for performance over time.

The author thanks the Institute for the Study of Business Markets for providing support for this research as a part of its Business Marketing Doctoral Support Award Competition; her dissertation chair, Murali Chandrashekaran, Bob Dwyer at the University of Cincinnati, Donald Lehmann at Columbia University, and the three anonymous JM reviewers for their helpful comments; and Jagdip Singh at Case Western Reserve University for his helpful comments and support.

( n1) The average age is based on 24 participants; 1 participant did not report her age. I also realized that study participants would be junior people in any organization. However, I do not have reason to believe that seniority of decision makers moderates or changes the relationship between dynamics in market knowledge and innovation. To check empirically whether this was the case in my data, I included the average work experience and age as main effects and interacted them with the variables in the innovation and performance model. I did not find any significant interactions; that is, the reported parameter estimates remain unchanged.

( n2) Although the ability to assess correctly the strength of competitors is an important aspect of competitor knowledge, attempts to collect data on the most influential competitor per segment yielded many missing data points and inaccuracies in recording (the demand on participants was significant). Respondents often forgot to circle the most influential competitor or, in assigning competing products to segments, noted that they were not sure in which segment the particular competitor's presence was the strongest. Because of the significant noise in the data, I chose not to include this measure in the analysis.

( n3) I also analyzed the data after weighing each team member's knowledge by distance from the team mean. This did not change the obtained substantive insights.

( n4) On the basis of H[sub1]-H[sub5], I specified X as X = [CUSTK, CUSTK x CUSTC, CUSTK x CUSTS, CUSTK x CUSTC x CUSTS, COMPK, COMPK x COMPC, COMPK x COMPS, COMPK x COMPC x COMPS, SAT, SAT², SIZE].

( n5) If the variance--covariance matrix of the disturbance term is a positive, definite matrix with unknown parameters that need to be estimated (e.g., cross-firm correlations, autocorrelations), generalized least squares is not feasible. Thus, a structure needs to be imposed on the model (i.e., obtain estimates of the unknown parameters and then proceed with the generalized least squares estimation). The resultant feasible generalized least squares estimates are consistent, unbiased, and asymptotically efficient.

( n6) I did not estimate a simultaneous system of equations because innovation takes place (and is measured) before subsequent performance, which is an important aspect of the study's longitudinal design. Empirically, correlation of the residuals from the two models (which account for contemporaneous effects and dependence on the past) is not significant (-.006, p < .588).

( n7) The effect of shared customer knowledge on innovation effort is given by b[sub3]CUSTK + b[sub4]CUSTK x CUSTC. I evaluated this at low (tenth percentile) and high (ninetieth percentile) values of CUSTK and CUSTC. Likewise, the effect of shared competitor knowledge on innovation effort is given by b[sub7]COMPK + b[sub8]COMPK x COMPC. I also evaluated this at low (tenth percentile) and high (ninetieth percentile) values of COMPK and COMPC.

( n8) The effect of change in customer knowledge on innovation effort is given by b[sub2]CUSTK + b[sub4]CUSTK x CUSTS. I evaluated this at low (tenth percentile) and high (ninetieth percentile) values of CUSTK and CUSTS. The effect of change in competitor knowledge on innovation effort is given by b[sub6]COMPK + b[sub8]COMPK x COMPS, which I also evaluated at low and high values of COMPK and COMPS.

( n9) The effect of customer knowledge level on innovation effort is given by b[sub1]+ b[sub2]CUSTC + b[sub3]CUSTS + b[sub4]CUSTK x CUSTS. I evaluated this at low (tenth percentile) and high (ninetieth percentile) values of CUSTC and CUSTS. The effect of competitor knowledge level on innovation effort is given by b5 + b[sub6]COMPC + b[sub7]COMPS + b[sub8]COMPK x COMPS, which I also evaluated at low and high values of COMPC and COMPS.

( n10) To address potential problems with multicollinearity, I created an instrument for innovation effort orthogonal to shared market knowledge, firm size, and strategic orientation.

( n11) Because size is a function of the previous period s performance, I constructed an instrumental variable estimate for size (based on a two-period lagged value) and performed the Hausman specification test. The results suggest that the estimated coefficients are not biased (χ[sub1d.f., sup2] = 1.78, n.s.).

Legend for Chart:

B - 1
C - 2
D - 3
E - 4
F - 5
G - 6
H - 7
I - 8
J - 9
K - 10
L - 11
M - 12
N - 13

               A                     B      C        D       E
                                     F      G        H       I
                                     J      K        L       M
                                                             N

1. Innovation effort                 1.00

2. Customer knowledge level          -.30   1.00

3. Change in customer knowledge       .12   -.17     1.00

4. Shared customer knowledge          .30    .28     -.12    1.00

5. Competitor knowledge level        -.06    .32      .33     .31
                                     1.00

6. Change in competitor knowledge     .17   -.02      .37    -.01
                                      .27   1.00

7. Shared competitor knowledge        .22   -.24      .27    -.21
                                      .20    .33     1.00

8. Satisfaction                      -.17    .33     -.19     .36
                                     -.13   -.15     -.30    1.00

9. Strategic orientation              .16   -.03     -.22     .02
                                     -.21    .01     -.05     .40
                                     1.00

10. Total shared market knowledge    -.17    .12     -.06     .01
                                      .15    .10      .04     .05
                                     -.20   1.00

11. Firm size                         .04    .30     -.15     .22
                                      .03   -.15     -.34     .41
                                      .21   -.18     1.00

12. Market share                     -.03    .23     -.14     .28
                                     -.15   -.10     -.30     .65
                                      .46   -.11      .57    1.00

13. Stock price index                 .01    .14     -.17     .13
                                     -.13   -.12     -.30     .53
                                      .39   -.23      .68     .92
                                                             1.00

Mean                                  613    .25      .63     .16
                                      .32    .13      .16    4.42
                                    14.57    .59   12,463   16.66
                                                             1324

Standard deviation                    859    .18      .47     .15
                                      .13    .10      .10    1.50
                                    13.44    .32     4366    8.19
                                                              638

Notes: Means for firm size and innovation effort are reported
in thousands of dollars.

Legend for Chart:

A - Assumption About Interdependence of Innovation
    Effort Across Firms
B - Assumption About Inertia in Innovation Effort
    No Intertia (ρ[subi] = 0 ∀ i)
C - Assumption About Inertia in Innovation Effort
    Similar Inertial Tendency Across Firms
    (λ[subi] = 0 ∀ i)
D - Assumption About Inertia in Innovation Effort
    Unique Inertial Tendency for Each Firm
    (ρ[subi] = ρ + λ[subi])

                 A                         B            C
                                                        D

No interdependence                      M[sub00]     M[sub01]
(σ[subij]= 0) ∀ i, j)                   M[sub02]

Interdependence σ[subij] free)    M[sub10]     M[sub11]
                                                     M[sub12]

Legend for Chart:

A - Conceptual Focus
B - Variables (Parameter)
C - Estimate (Standard Error)

        A
                    B                            C

Customer knowledge diffusion
   CUSTK (β[sub1])                     -.952(**)
                                            (.437)

   CUSTK x CUSTC (β[sub2])             -.229
                                            (.366)

   CUSTK x CUSTS (β[sub3])             1.196(***)
                                            (.306)

   CUSTK x CUSTC x CUSTS (β[sub4])     3.072(***)
                                            (.370)

Competitor knowledge diffusion
   COMPK (β[sub5])                     -.569
                                            (.316)

   COMPK x COMPC (β[sub6])              .315
                                            (.749)

   COMPK x COMPS (β[sub7])              .552(*)
                                            (.316)

   COMPK x COMPC x COMPS (β[sub8])     3.664(**)
                                           (1.729)

Feedback from marketplace: effect of satisfaction
   SAT (β[sub9])                     -11.50(***)
                                           (1.88)

   SAT x SAT (β[sub10])                 .205(***)
                                            (.029)

Firm size
   SIZE (β[sub11])                     -.883
                                           (1.14)

(*) p < .10.

(**) p < .05.

(***) p < .0001.

A. Net Effect of Shared Customer/Competitor Knowledge

Change in Customer Knowledge

Legend for Chart:

B - Low
C - High

 A           B             C

Customer Knowledge Level

Low        .724(**)     12.31(**)
          (.123)        (2.088)

High      2.936(**)     49.91(**)
          (.308)        (5.229)

Change in Competitor Knowledge

Legend for Chart:

B - Low
C - High

 A           B             C

Customer Knowledge Level

Low        .735(**)     2.274(**)
          (.254)        (.598)

High      3.970(*)      12.28(**)
         (1.370)        (3.230)

B. Net Effect of Change in Customer/Competitor Knowledge

Shared Customer Knowledge

Legend for Chart:

B - Low
C - High

 A           B             C

Customer Knowledge Level

Low        .154          4.824(**)
          (.120)         (.463)

High      2.621         82.01(**)
         (2.038)        (7.867)


Shared Competitor Knowledge

Legend for Chart:

B - Low
C - High

 A           B             C

Customer Knowledge Level

Low        .105(*)       1.131(*)
          (.059)         (.475)

High       .566(*)       6.105(*)
          (.320)        (2.564)

C. Net Effect of Customer/Competitor Knowledge Level

Shared Customer Knowledge

Legend for Chart:

B - Low
C - High

 A           B             C

Change in Customer Knowledge

Low       -.636          6.243(**)
          (.416)        (1.416)

High       .058         27.95(**)
          (.465)        (2.817)

Shared Competitor Knowledge

Legend for Chart:

B - Low
C - High

 A           B             C

Change in Competitor Knowledge

Low       -.406          1.625(*)
          (.354)         (.8.39

High       .034          6.401(**)
          (.277)        (1.596)

(*) p < .05.

(**) p < .0001.

Notes: All entries are estimates. Standard errors
are in parentheses.

Legend for Chart:

A - Variable (Parameter)
B - Main Study Performance Measured as Market Share
C - Main Study Performance Measured as Stock Price Index
D - Validation Study 2 Performance Measured as Market Share

           A
                       B             C               D

Innovation effort (β[sub1])
                      .090         .027(**)         .008
                    (1.125)       (.005)           (.009)

Innovation effort x shared
market knowledge (β[sub2])
                     5.438(*)       .017(**)        .048(**)
                     (.866)        (.003)          (.015)

Innovation effort x firm size (β[sub3])
                     -.231(**)     -.003(**)       -.002(**)
                     (.082)        (.0004)         (.0003)

Firm size (β[sub4])
                      .921(**)      .806(**)        .247(**)
                     (.437)        (.046)          (.055)

Shared market knowledge (β[sub5])
                     -.590        -1.078(**)      -2.350(**)
                     (.607)        (.302)          (.767)

Strategic orientation (β[sub6])
                      .121(**)      .093(**)        .044(**)
                     (.022)        (.009)          (.008)

(*) p < .05.

(**) p < .0001.

Notes: Entries are parameter estimates from the
corresponding model. Standard errors are in parentheses.

Legend for Chart:

A - Aspects of Study
B - Main Study
C - Pretest 1
D - Validation Study 1
E - Validation Study 2

      A
                   B                           C
                   D                           E

Microlevel data collected
          Knowledge diffusion,        Satisfaction with past
          satisfaction with past      performance, group
          performance, strategic      dynamics, information
          orientation                 use issues

          None                        Satisfaction with
                                      past performance,
                                      group dynamics,
                                      strategic orientation

Number of firms (teams)
          6                           5

          5                           6

Number of time periods
          8                           6

          9                           9

Analysis data (number of
firm/time periods)
          48                          30

          45                          54

Support for the partial
adjustment model?
          Yes                         Yes
          (χ[sub1, sup2]          (χ[sub1, sup2]
          = 17.383,                   = 15.122,
          p < .0001)                  p < .0001)

          Yes                         Yes
          (χ[sub1, sup2]          (χ[sub1, sup2]
          = 26.698,                   = 15.07,
          p < .0001)                  p < .0001)

Average value (standard
deviation) of ρ
          .732                        .753
         (.188)                      (.189)

          .763                        .700
         (.138)                      (.116)

Is there performance
interdependence
across firms?
          Yes                         Yes
          (χ[sub15, sup2]         (χ[sub10, sup2]
          = 36.148,                   = 23.71,
          p < .01)                    p < .01)

          Yes                         Yes
          (χ[sub10, sup2]         (χ[sub10, sup2]
          = 74.80,                    = 35.46,
          p < .01)                    p < .01)

Legend for Chart:

A - Criteria
B - Hypothesized Model Performance Main Study
C - Hypothesized Model Performance Validation Study 2
D - Hypothesized Model Performance Validation Study 3

              A                     B        C         D

Correlation between actual          .80      .64       .92
and predicted performance

Mean absolute error                2.52     2.40      2.59

Mean absolute percentage error     .175     .178       .22

Mean square error                 11.43     9.10     10.19

Theil's U                          .092     .087      .071

DIAGRAM: FIGURE 1 Conceptual Framework

Ansoff, H. Igor (1984), Implanting Strategic Management. Englewood Cliffs, NJ: Prentice Hall.

Argote, Linda (1999), Organizational Learning: Creating, Retaining, and Transferring Knowledge Norwell, MA: Kluwer Academic Publishers.

Argyris, Chris and Donald A. Schön (1978), Organizational Learning: A Theory of Action Perspective Reading, MA: Addison-Wesley.

Audia, Pino G., Edwin A. Locke, and Ken G. Smith (2000), "The Paradox of Success: An Archival and Laboratory Study of Strategic Persistence Following a Radical Environmental Change," Academy of Management Journal, 43 (October), 837-53.

Ayers, Doug, Robert Dahlstrom, and Steven Skinner (1997), "An Exploratory Investigation of Organizational Antecedents to New Product Success," Journal of Marketing Research, 34 (February), 107-116.

Baldridge, J. Victor and Robert A. Burnham (1975), "Organizational Innovation: Individual, Organizational, and Environmental Impacts," Administrative Science Quarterly, 20 (June), 165-76.

Barney, Jay (1991), "Firm Resources and Sustained Competitive Advantage," Journal of Management, 17 (March), 99 120.

Bayus, Barry, Gary Erickson, and Robert Jacobson (2003), "The Financial Rewards of New Product Introductions in the Personal Computer Industry," Management Science, 49 (February), 197-210.

Bertels, T. and C. Savage (1998), "Understanding Knowledge in Organizations," in Knowing in Firms: Understanding, Managing, and Measuring Knowledge G. von Krogh, J. Roos, and D. Kleine, eds. London: Sage Publications.

Boulding, William (1990), "Unobservable Effects and Business Performance: Do Fixed Effects Matter?" Marketing Science, 9 (Winter), 88-91.

Cannon-Bowers, J.A., E. Salas, and S. Converse (1993), "Shared Mental Models in Expert Team Decision Making," in Individual and Group Decision Making, N. John Castellan Jr., ed. Hillsdale, NJ: Lawrence Erlbaum Associates, 221-46.

Carlile, Paul R. (2002), "A Pragmatic View of Knowledge and Boundaries: Boundary Objects in New Product Development," Organization Science, 13 (July August), 442-55.

Chan, Kim and Renee Mauborgne (1997), "Fair Process: Managing in the Knowledge Economy," Harvard Business Review, 75 (July-August), 65-75.

Chandrashekaran, Murali, Raj Mehta, Rajesh Chandrashekaran, and Rajdeep Grewal (1999), "Market Motives, Distinctive Capabilities, and Domestic Inertia: A Hybrid Process Model of Innovation Generation," Journal of Marketing Research, 36 (February), 95-112.

Chandy, Rajesh K. and Gerard J. Tellis (2000), "The Incumbent's Curse? Incumbency, Size, and Radical Product Innovation," Journal of Marketing, 64 (July), 1-17.

Cheng, Leonard and Zhigang Tao (1999), "The Impact of Public Policies on Innovation and Imitation: The Role of R&D Technology in Growth Models," International Economic Review, 40 (February), 187-207.

Clark, Bruce and David B. Montgomery (1998), "Deterrence, Reputations, and Competitive Cognition," Management Science 44 (January), 62-82.

----- and ----- (1999), "Managerial Identification of Competitors," Journal of Marketing, 63 (July), 67-83.

Cohen, Wesley and Steven Klepper (1996a), "Firm Size and the Nature of Innovation Within Industries: The Case of Process and Product R&D," The Review of Economics and Statistics 78 (March), 232-43.

----- and ----- (1996b), "A Reprise of Size and R&D," The Economic Journal, 106 (July), 925-51.

----- and Daniel Levinthal ((1990), "Absorptive Capacity: A New Perspective on Learning and Innovation," Administrative Science Quarterly, 35 (March), 128-52.

Coleman, James (1968), "The Mathematical Study of Change," in Methodology in Social Research, H.M. Blalock and A. Blalock, eds. New York: McGraw-Hill, 428-78.

Cyert, Richard M. and James G. March (1963), A Behavioral Theory of the Firm Englewood Cliffs, NJ: Prentice Hall.

Day, George (1994), "The Capabilities of Market-Driven Organizations," Journal of Marketing 58 (October), 37-52.

----- (1997), "Maintaining the Competitive Edge: Creating and Sustaining Advantages in Dynamic Competitive Environments," in Wharton on Dynamic Competitive Strategy, George Day and David Reibstein, eds. New York: John Wiley & Sons.

----- and Prakash Nedungadi ((1994), "Managerial Representations of Competitive Advantage," Journal of Marketing, 58 (April), 31-44.

Dickson, Peter (1992), "Toward a General Theory of Competitive Rationality," Journal of Marketing 56 (January), 69-83.

----- (1996), "The Static and Dynamic Mechanics of Competition: A Comment on Hunt and Morgan's Comparative Advantage Theory," Journal of Marketing, 60 (October), 102-106.

Dosi, Giovanni (1988), "Sources, Procedures, and Microeconomic Effects of Innovation," Journal of Economic Literature, 26 (September), 1120-71.

Frank, Kenneth A. and Kyle Fahrbach (1999), "Organization Culture as a Complex System: Balance and Information in Models of Influence and Selection," Organization Science, 10 (May June), 253 77.

Gates, William H. (1999), Business at the Speed of Thought New York: Warner Books.

Gatignon, Hubert and Jean-Marc Xuereb (1997), "Strategic Orientation of the Firm and New Product Performance," Journal of Marketing Research, 34 (February), 77-90.

Geroski, Paul, Steve Machin, and John Van Reenen (1993), "The Profitability of Innovating Firms," RAND Journal of Economics 24 (Summer), 198-211.

Glazer, Rashi, Joel Steckel, and Russell Winer (1989), "The Formation of Key Marketing Variable Expectations and Their Impact on Firm Performance: Some Experimental Evidence," Marketing Science, 8 (Winter), 18-34.

-----, -----, and ----- (1990), "Judgmental Forecasts in a Competitive Environment: Rational Vs. Adaptive Expectations," International Journal of Forecasting, 6 (July), 149-62.

-----, -----, and ----- (1992), "Locally Rational Decision Making: The Distracting Effect of Information on Managerial Performance," Management Science, 38 (February), 212 26. Grandori, Anna and Bruce Kogut (2002), "Dialogue on Organization and Knowledge," Organization Science, 13 (May-June), 224-31.

Greene, William H. (2002), Econometric Analysis New York: Macmillan.

Greve, Henrich R. (1999), "The Effect of Core Change on Performance: Inertia and Regression Toward the Mean," Administrative Science Quarterly, 44 (September), 590-614.

----- (2002), "Sticky Aspirations: Organizational Time Perspective and Competitiveness," Organization Science, 13 (January-February), 1-17.

Hammond, Kenneth and David Summers (1972), "Cognitive Control," Psychological Review, 79 (January), 58-67.

Hargadon, Andrew and Angelo Fanelli (2002), "Action and Possibility: Reconciling Dual Perspectives of Knowledge in Organizations," Organization Science, 13 (May-June), 290-302.

Hoopes, David and Steven Postrel (1999), "Shared Knowledge, 'Glitches,' and Product Development Performance," Strategic Management Journal, 20 (September), 837-65.

Huber, G. (1991), "Organizational Learning: The Contributing Processes and the Literatures," Organization Science, 2 (February), 88-115.

Hunt, Shelby D. and Robert M. Morgan (1995), "The Comparative Advantage Theory of Competition," Journal of Marketing, 59 (April), 1-15.

Isen, Alice M. and R.A. Baron (1991), "Positive Affect as a Factor in Organizational Behavior," in Research in Organizational Behavior, Vol. 13, L.L. Cummings and B.M. Staw, eds. Greenwich, CT: JAI Press, 1-53.

Ittner, Christopher and David Larcker (1997), "Product Development Cycle Time and Organizational Performance," Journal of Marketing Research, 34 (February), 13-23.

Jacobson, Robert (1990), "Unobservable Effects and Business Performance," Marketing Science, 9 (Winter), 74-85.

Kessler, Eric H. and Alok K. Chakrabarti (1996), "Innovation Speed: A Conceptual Model of Context, Antecedents, and Outcomes," Academy of Management Review, 21 (October), 1143-91.

Kotler, Philip (2000), Marketing Management Englewood Cliffs, NJ: Prentice Hall.

Larreche, Jean-Claude and Hubert Gatignon (1999), Markstrat3: Instructor's Manual Redwood City, CA: The Scientific Press.

Lee, Jeho (2003), "Innovation and Strategic Divergence: An Empirical Study of the U.S. Pharmaceutical Industry from 1920 to 1960," Management Science, 49 (February), 143-59.

Leeflang, Peter S.H., Dick R. Wittink, Michel Wedel, and Philippe A. Naert (2000), Building Models for Marketing Decisions Boston: Kluwer Academic Publishers.

March, James G. and Robert I. Sutton (1997), "Organizational Performance as a Dependent Variable," Organization Science 8 (November-December), 698-706.

McKee, Daryl, P. Rajan Varadarajan, and William Pride (1989), "Strategic Adaptability and Firm Performance: A Market-Contingent Perspective," Journal of Marketing, 53 (July), 21-35.

Miller, Chet C., Linda M. Burke, and William H. Glick (1998), "Cognitive Diversity Among Upper-Echelon Executives: Implications for Strategic Decision Processes," Strategic Management Journal, 19 (January), 39-58.

Miller, Danny and Ming-Jer Chen (1994), Sources and Consequences of Competitive Inertia: A Study of the U.S. Airline Industry," Administrative Science Quarterly, 39 (March), 1-23.

Miyagiwa, Kaz and Yaku Ohno (1999), "Credibility of Protection and Incentives to Innovate," International Economic Review 40 (February), 143-63.

Moorman, Christine and Anne Miner (1997), "The Impact of Organizational Memory on New Product Performance and Creativity," Journal of Marketing Research, 34 (February), 91-106.

Oliver, Richard L. (1980), "A Cognitive Model of the Antecedents and Consequences of Satisfaction Decisions," Journal of Marketing Research, 17 (November), 460-69.

Polanyi, Michael (1967), The Tacit Dimension Garden City, NY: Doubleday.

Porter, Michael E. (1995), "Toward a Dynamic Theory of Strategy," in Fundamental Issues in Strategy: A Research Agenda, R.P. Rumelt, D.E. Schendel, and D.J. Teece, eds. Boston: Harvard Business School Press, 423-62.

Schumpeter, Joseph A. (1942), Capitalism, Socialism, and Democracy New York: Harper and Brothers.

Senge, Peter (1993), The Fifth Discipline: The Art and Practice of the Learning Organization London: Century Business.

----- (1997), "Communities of Leaders and Learners," Harvard Business Review 75 (September-October), 30-32.

Simon, Herbert A. (1991), "Bounded Rationality and Organizational Learning," Organization Science, 2 (January), 125-34.

Spender, J.C. (1996), "Making Knowledge the Basis of a Dynamic Theory of the Firm: Overview," Strategic Management Journal" 17 (Winter Special Issue), 45-63.

Sternberg, Robert J. (1996), Cognitive Psychology Fort Worth, TX: Harcourt Brace.

Sutcliffe, Kathleen M. and Klaus Weber (2003), "The High Cost of Accurate Knowledge," Harvard Business Review, 81 (May), 74-82.

Szymanski, David M., Sundar G. Bharadwaj, and P. Rajan Varadarajan (1993), "An Analysis of the Market Share --Profitability Relationship," Journal of Marketing, 57 (July), 1-18.

Teece, David J. (2001), "Strategies for Managing Knowledge Assets: The Role of Firm Structure and Industrial Context," in Managing Industrial Knowledge: Creation, Transfer, and Utilization Ikujiro Nonaka and David J. Teece, eds. London: Sage Publications, 125-44.

-----, Gary Pisano, and Amy Shuen (1997), "Dynamic Capabilities and Strategic Management," Strategic Management Journal 18 (August), 509.

Van Bruggen, Gerrit H., Ale Smidts, and Berend Wierenga (1998), "Improving Decision Making by Means of a Marketing Decision Support System," Management Science, 44 (May), 645-58.

von Hayek, Friedrich (1989), "Spontaneous ('Grown') Order and Organized ('Made') Order," in Order--With or Without Design Naomi Modlofsky, ed. London: Center for Research into Communist Economies, 101-123.

Voss, Glenn B. and Zannie G. Voss (2000), "Strategic Orientation and Firm Performance in an Artistic Environment," Journal of Marketing, 64 (January), 67-83.

Webster, Frederick (1997), "The Future Role of Marketing in the Organization," in Reflections on the Futures of Marketing, Donald R. Lehmann and Katherine E. Jocz, eds. Cambridge, MA: Marketing Science Institute, 39-67.

The partial adjustment process can be expressed as follows (see Coleman 1968; Greve 1999):

(A1) ∂PERF[subit]/∂t = r[subi](PERF[subit, sup*] - PERF[subit],

where PERF[subit] and PERF[subit, sup*] denote the realized and potential performance for the ith firm at time t, and r[subi] captures the adjustment rate for the ith firm. When r[subi] = 0, there is no adjustment, and a condition of complete inertia is in evidence. When r[subi] is large, there is evidence of little inertia. Solving Equation A1 yields the following solution:

(A2) ln [PERF[subit, sup*] - PERF[subit]/PERF[subi,t - 1, sup*] - PERF[subi,t - 1]] = -r[subi],

which simplifies to

(A3) PERF[subit] = PERF[subit, sup*] - e[sup-r[subi]](PERF[subi,t - 1, sup*] - PERF[subi,t - 1).

Substituting ρ[subi] = e[sup-r[subi]], I obtain

(A4) PERF[subit] = PERF[subit, sup*] - ρ[subi] (PERF[subi,t - 1, sup*] - PERF[subi,t - 1).

Finally, specifying PERF[subit, sup*] = X[subit]β + ε[subit], I express the autoregressive partial-adjustment model as follows:

(A5) PERF[subit] = X[subit]β - ρ(X[subi,t-1]β - PERF[subi,t - 1] + ε[subit] - ρ[subi]ε[subi,t - 1],

which can be expressed as follows:

(A6) PERF[subit] = X[subit]β - ρ[subi](X[subi,t - 1]β - PERF[subi,t - 1] + η[subit,

where η[subit] = ε[subit] - ρ[subi]ε[subi,t - 1].

~~~~~~~~

By Detelina Marinova

Detelina Marinova is Assistant Professor of Marketing, Weatherhead School of Management, Case Western Reserve University (e-mail: detelina.marinova@case.edu).

An Accidental Venture into Academics 

In the summer of 2002, Terry Clark asked me to write an essay about the history and development of my professional activities. Although I was flattered, I also realized that this must be a sign that I had reached a mature, sometimes confused with wise, stage of my life cycle. I delayed informing others about my maturity by waiting a year to complete this essay. However, unless the biomedical innovations with regard to aging accelerate rapidly, I need to be realistic: At this point, I still have the time and wherewithal to ponder new options. Besides, the process of reflection might be instructive.

The title of my essay is intended to convey that seemingly random events have exerted major influences in my life. It is well known that we have little control over many aspects of our lives. Our diverse genetic makeup is fixed to a large degree, and we face extensive heterogeneity in the opportunities available to us. Nevertheless, there is a lot we can do to shape our lives. I will attempt to illustrate by reviewing my own state of affairs.

I grew up in a family and a culture dominated by a Calvinistic work ethic. Family life was dominated by my father's bakery, in part because the production facility was in the back of the house and the retail outlet in the front. All family members were expected to assist with the business when needed; therefore, demands on my time could occur any day I returned from school. To minimize the chance, I often joined friends after school to spend time at their homes or to play soccer. Broader family gatherings were important, and in the presence of extended family members, my father seemed intent on demonstrating that he had achieved at least as much financial success as two uncles with university degrees had.

Although neither of my parents had gone to high school, they seemed to believe that education would be critical to my future success. In the Netherlands, the most selective secondary school option was required for entry into universities. A second option was intended to provide the basis for low-and midlevel professional careers and a third one for skilled labor. The Department of Education mandated that all primary school children be evaluated during their last year with respect to their potential for the three options. The counselor who saw me was probably conscious of social classes, and my parents' modest education may have influenced him to recommend that I attend the school for skilled labor. However, he added that if my parents felt strongly about it, the second option might be suitable. Because neither of these options was consistent with my grades, I give my parents credit for ignoring this advice. Their decision to send me to the highest level of secondary education was especially important given that I had no interest in succeeding my father in his business. The uncertainties of a small business presented obvious pitfalls, and I vowed that my own family would enjoy a home life separate from business.

Toward the end of my secondary education, I contemplated studying economics. University study, heavily subsidized by the state, was a risky activity in the 1960s. About half the students in Holland failed to continue after the first or second year, and most graduates took longer than the norm of five to six years for the equivalent of a master's degree.( n1) I decided to pursue an unconventional option instead and applied for admission to Nyenrode, a private college created after World War II to prepare students for careers in international business, with financial support from companies such as Philips, Shell, and Unilever. Course content placed heavy emphasis on macro-and microeconomics and foreign languages including English, French, German, Spanish, and Swedish. There were also ample opportunities for participation in a wide range of sports and other activities designed to enhance social skills.

A week before graduation from Nyenrode, I was invited to return to campus for an interview related to a scholarship for study abroad. I was planning to stay in the Netherlands to pursue a career in accounting. Because accounting conventions differed widely between countries, I argued that the scholarship was irrelevant to me. However, on graduation day, the college president informed me that I had won the scholarship on the basis of my exam scores. He asked that I simply accept the prize during the ceremonies and decide later whether I would actually use it. He knew I was due for up to two years of compulsory Dutch military service and would have ample time to consider it.

During a one-year officer-training program, I discussed professional options with my fiancée and with fellow trainees who were mostly law and economics graduates. Everyone agreed that it would be a mistake for me not to accept the scholarship. My fiancíe and I decided we should complete a year of marriage before we embarked on a foreign venture. She initially favored European options, such as graduate study in economics at Sankt Gallen in Switzerland. However, America beckoned.

Although a scholarship made it easier to justify going abroad, the actual benefits were still unclear to me. Nevertheless, I accepted admission to a program at the University of Oregon. Because marketing was considered especially strong in the United States, I abandoned my interest in accounting and chose to concentrate instead in marketing. During my first quarter at Oregon, I took a course in statistics from James Reinmuth, who made complex material understandable and interesting. I ended up taking many more statistics courses and far fewer marketing courses than I had initially intended during two years of study at Oregon. My plan was to return to Holland or at least Europe, perhaps to work in market research for Philips or Unilever or to join a consulting firm. Just when I began contemplating options, two things interfered.

First, my wife Marian now preferred to stay in the United States. I might have persuaded her to return to Europe, but because she had joined me reluctantly when I chose the United States over Switzerland, it did not seem fair for me to ask her to move in both directions across the Atlantic against her preferences. She had worked full-time while I studied, and her interests deserved full consideration in our decisions. Before we married, her mother asked me whether I was sure Marian was the "right one" for me. I must have had a mind for statistics because I remember contemplating something along the lines of the following: "Given the large number of unobservables, how could I know if she is the one? Of all the eligible females, I knew just a handful well enough to make somewhat informed judgments about their suitability." I said something like that to my future mother-in-law. However, I also emphasized that I thought Marian and I were quite compatible and that I was committed to building a strong relationship. We had been childhood sweethearts and had shown a strong interest in each other at various stages in our lives. I had read a book or two about matrimony and had formed some tentative opinions on the characteristics of successful and happy marriages. Happy marriages seemed to depend especially on the compatibility of partners but also on a great deal of personal attention and flexibility by both partners. With this philosophy in mind, it seemed that it was my turn to be flexible. Thus, staying in the United States became a priority.

Second, Reinmuth suggested that I consider doctoral study. Although the thought had not crossed my mind, the idea was sufficiently intriguing for me first to try full-time teaching. I obtained one offer at a public college in Washington State, where I taught mostly statistics for two years. Research had fascinated me already. For example, I did a laboratory experiment with a fellow student on odd pricing, and Reinmuth proposed to combine a project I had completed with his own research, which several years later resulted in my first publication (Reinmuth and Wittink 1974).

I was especially intrigued by the doctoral programs at Northwestern and Purdue. I chose Purdue because it seemed to fit better with my interests. Still, during the second year of the three-year program, I had doubts about my ability to complete it. My wife convinced me to continue (or perhaps she said that quitting was unacceptable). Our daughter, who was two years old when I started the doctoral program, also played an important role. She would often accompany me during the weekends on visits to the computer center where I did batch processing of programs. She enjoyed playing with IBM punch cards and drawing on computer output.

Purdue's Krannert School had achieved a strong reputation because of Frank Bass and Mike Pessemier. In the fall of 1974, my cohort received one or more offers from a variety of schools, including Berkeley, Carnegie, Chicago, Harvard, MIT, Northwestern, Stanford, and Wharton. I accepted the offer from Stanford partly because of its research excellence across fields and the presence of marketing scholars such as Dave Montgomery, Mike Ray, Seenu Srinivasan, and Peter Wright. Interestingly, although I would join the marketing group, my job was actually to teach statistics and econometrics. Marian gave birth to our second child at about the same time I defended my thesis.

Stanford

During the doctoral program, much of my research interest focused on econometric models, including the latest methods for pooling cross-sectional and time-series data (Bass and Wittink 1975), and I pursued interaction effects between price and advertising (Wittink 1977; see also Kaul and Wittink 1995). At Stanford, Dave Montgomery and Seenu Srinivasan pursued customer preference (conjoint) measurement. This emerging research activity offered interesting opportunities. At the time, little was known about differences in performance between, for example, alternative methods (Wittink and Cattin 1981) or external validity (Montgomery and Wittink 1980). The 1981 paper I coauthored with Philippe Cattin was initially rejected by Journal of Marketing Research (JMR) in 1976. It remained on the shelf for four years until a colleague suggested that we resubmit it to the same journal!

An important mentor of mine was Richard Johnson. My first meeting with Rich was at Market Facts in Chicago, where he had developed an attribute trade-off approach that focused on two attributes at a time (Johnson 1974). He had an application that involved characteristics of life partners to interest potential users in conjoint applications.

Other research projects included a survey of commercial practices (Cattin and Wittink 1982) and an application to the marketing of arts programs (Currim, Weinberg, and Wittink 1981). The design of the latter study had three attributes with three levels and three attributes with two levels. The empirical results suggested a strong artificial effect that I focused on in various projects (e.g., Steenkamp and Wittink 1994; Wittink, Krishnamurthi, and Nutter 1982; Wittink, Krishnamurthi, and Reibstein 1990; Verlegh, Schifferstein, and Wittink 2002).

I fondly recall making a presentation at the first Advanced Research Techniques Forum on the number-of-attribute-levels effect that at that time had been shown to exist only for full-profile and trade-off matrix designs (Wittink 1991). Joel Huber asked publicly if I thought the effect would apply to adaptive conjoint analysis (ACA) (Johnson 1987). In some sense, ACA was based on Johnson's tradeoff matrix approach; it customizes trade-off questions based on self-explicated data. I suggested that the number-of attribute-levels effect might not apply to ACA. Joel suggested that it would. After my presentation, two practitioners offered to join us for a study. We designed an experimental study with four treatments, each of which had two options (see Huber et al. 1993). The empirical results for the number-of-levels effect showed that Joel was correct: ACA also showed a number-of-levels effect (Wittink et al. 1992), but its magnitude was much smaller than it was for full profile. Interestingly, we also found that the effect did not need to occur in ACA. In subsequent versions of ACA, Johnson modified the integration of self-explicated data with paired preference intensity judgments, and this modification reduced the number-of-levels effect further (Orme 1998).

An important element of my time at Stanford consisted of the development of an econometric model to aid the MBA student admission process. The idea was to relate academic performance in the MBA program to many predictors, including GMAT scores, undergraduate grade point average, and a measure of the quality of the undergraduate institution. The project spanned multiple years and included the difficulty of obtaining cooperation from successive admission directors. A working paper (never submitted for publication) discusses important issues regarding measurement, model specification, estimation, and validation (Srinivasan, Wittink, and Zweig 1981). We proposed a two-stage process for admission decisions. In the first stage, a model-based prediction of each applicant's probability of failing to satisfy minimum requirements in the quantitative and/or the managerial components in the core MBA program is obtained. Applicants with high failure likelihoods would be eliminated. In the second stage, the admission director could focus primarily on the remaining candidates' career potentials. Seenu was the primary architect, and it was a privilege for me to work with him on this project.

My Stanford period also included the Market Measurement and Analysis Conference that Dave Montgomery and I organized in 1979 on a suggestion by Frank Bass. We proposed to have select academics and practitioners interact about research. This conference is now considered the first Marketing Science conference (see Wittink 2001).

During my sixth year at Stanford, Cornell offered me a position with tenure. Marian and I decided that upstate New York is an idyllic environment in which families thrive. Another rationalization for the move was that we believed that our two children would benefit from exposure to the seasons, and Holland would be easier to visit. I am sure that the uncertainty of tenure at Stanford played no role, yet I do recall a discussion with junior-faculty colleagues at Stanford about prospects for tenure, partly as a result of the frequent failures of internal candidates to be promoted. A non-marketing colleague told me that if I just ignored my family for the remaining time, tenure would be attainable. We moved to Ithaca in 1981.

Cornell

For various reasons, including my acceptance of administrative responsibilities, and the onset of severe medical problems, my research output declined during the first part of my time at Cornell. A sabbatical leave in 1988-89, spent at Northwestern's Kellogg School and at the Faculty of Economics of the University of Groningen in the Netherlands, was critical to the reinvigoration of my research activities. My visit in Holland showed that Peter Leeflang had created a research culture that rivaled the best in the United States, and we embarked on various projects, including competitive reaction research (Leeflang and Wittink 1992, 1996, 2001). The reemergence of my interest in research was also facilitated by the availability of scanner data. In the mid-1980s, I accepted an invitation from ACNielsen to work on the development of a price-promotion model based on weekly store-level scanner data. My work at ACNielsen culminated in a paper submitted to JMR that describes the initial promotion model along with empirical results. Rejection of the SCAN*PRO paper (Wittink et al. 1988) by JMR was swift. Because there was no new estimation methodology, the reviewers argued that the paper was not suitable for the journal. Still, the model appears to satisfy Little's (1970) criteria for structure: It is simple, complete, adaptive, and robust. I regret that my coauthors and I never sent the paper elsewhere. Quite a few years later, Don Morrison, who had nothing to do with the rejection at JMR, provided some redemption when he opined that this paper should have been published (for research that uses empirical results in the working paper, see Morrison and Silva-Rossa 1995).

The SCAN*PRO model provided a basis for further research. For example, Christen and colleagues (1997) show the bias in a model of linearly aggregated data when the model of disaggregate data is nonlinear. Foekens, Leeflang, and Wittink (1999) show how promotion effects depend on the frequency and intensity of previous price discounts. The inclusion of dynamic effects is also a way to address the Lucas (1976) critique. Van Heerde, Leeflang, and Wittink (2000) show that cross-period effects can be obtained in models estimated from store data. Van Heerde, Leeflang, and Wittink (2001) use semiparametric estimation of promotion effects to show that these effects have complex main (and crossover) interaction effects. They also demonstrate superior predictive validity of the semiparametric approach relative to traditional approaches. Van Heerde, Leeflang, and Wittink (2002) describe the evolution of research related to SCAN*PRO.

I am particularly excited about two recent articles that build on the innovative and award-winning research by Sunil Gupta (1988). Much of the development in these two articles is due to the highly capable work of Harald van Heerde, a highly competent econometrician by training and a terrific research colleague. In one article, Van Heerde, Sachin Gupta, and Wittink (2003) transform the well-known household data-based elasticity decomposition result, which has been misinterpreted as reporting that the vast majority of the sales effect is due to brand switching, into a unit-sales decomposition. In the other article, Van Heerde, Leeflang, and Wittink (2005) show that substantively meaningful unit-sales decompositions can also be obtained from store data. The framework in the latter article is extremely powerful and attractive.

My association with Peter Leeflang continues to be productive. We completed a book on model building (Leeflang et al. 2001), a revision of the highly regarded book he and Philippe Naert completed when quantitative model building in marketing was in its infancy (Naert and Leeflang 1978). My research association with Peter was formalized with an honorary faculty appointment at the Department of Economics of the University of Groningen in 1993, and I received a great accolade for my research when I was elected to the esteemed Royal Dutch Academy of Sciences in 2001.

In 1990, Cornell had the privilege of attracting Pradeep Chintagunta to join the faculty. Pradeep is highly capable, enormously productive, and a superb colleague. His departure for the University of Chicago in the mid-1990s was a severe professional and personal loss for me. Both of us worked with Cornell's strong cadre of doctoral students, and I want to acknowledge especially Sachin Gupta, with whom both of us worked extensively (Gupta, Chintagunta, and Wittink 1997; Gupta et al. 1996).

In the mid-1990s, I began working with Ed McLaughlin at Cornell's Department of Agricultural Economics. This department has a strong marketing component with extensive contacts in several industries, especially the food industry. Ed and I have worked with associates on customer satisfaction research (Gomez, McLaughlin, and Wittink 2004; Sirohi, McLaughlin, and Wittink 1998), and we are now extending the research with enhanced data from the food industry. Another association that is important to me is that with my fellow doctoral student at Purdue. Dave Reibstein and I worked together on conjoint analysis (Wittink et al. 1989; Wittink, Krishnamurthi, and Reibstein 1990) and strategy (Keil, Reibstein, and Wittink 2001). In 2001, we co-organized a special-interest conference on competition, sponsored by the Marketing Science Institute, that will lead to a special issue of Marketing Science.

I want to acknowledge the role of Tom Dyckman during my Cornell period. Tom was a terrific mentor, and I am deeply grateful for his support. I take great pride in the fact that Cornell had a much stronger and larger group of marketing faculty in the late 1990s than it had in the early 1980s, when it was known primarily for accounting and finance. BusinessWeek (1998) placed Cornell in the top ten only in marketing. So why move?

Yale

Yale is attractive for many reasons, including a strong research culture at the School of Management. The transition from Cornell was amazingly smooth, and I felt immediately comfortable. Ravi Dhar and Subrata Sen helped facilitate my integration into the Yale culture. Importantly, we have been able to create a superb group of researchers with the additions of On Amir, Dina Mayzlin, Nathan Novemsky, and K. Sudhir. Another benefit is the proximity to New York City. I am a partner at Blue Flame Data Inc., a firm that I cofounded with Sev Keil (see Wittink and Keil 2000) that is located in Manhattan, and my wife and I usually have season tickets to the Metropolitan Opera.

My move to Yale in 1998 also spurred research in health care. Patients demand a greater role in the decision-making process for chronic diseases. Yet physicians do not have the time, the inclination, the ability, or the preparation to learn each patient's unique perspectives on alternative treatments that differ on multiple criteria. My personal experience illustrates the problem. I was diagnosed with type 1 (insulin-dependent) diabetes in 1983. My family physician presented multiple treatment options, such as daily injections and an insulin pump. The latter requires the insertion of a subcutaneous device attached to a visible beeperlike instrument, a method that seemed excessively obtrusive to me at the time (I had visions of being a patient walking in a hospital with an IV bottle attached to a transportable stand). I now recognize that this perception was excessively negative partly because the autoimmune disease diagnosis depressed me, and I preferred to hide the diabetic condition. However, knowledge of the severe consequences of hyperglycemia, such as cardiovascular disease and neuropathy, made it impossible for me to ignore the disease. So I chose insulin injections to cover basal requirements and the expected carbohydrate intake for 24 hours at a time. Choosing this option constrained my food intake in the sense that I risked hypoglycemic reactions if I did not eat at specified times. My interest in minimizing the chances of severe long-term effects from persistent hyperglycemia made me favor accepting the short-term risks of hypoglycemia.

I mention this for several reasons. One is that having type 1 diabetes affected my life enormously. Coming to grips with being diabetic reduced my research activity. My wife and children were greatly affected by the multiple hypoglycemic bouts I experienced. I became unconscious several times and frequently experienced reduced cognitive abilities. An obvious solution to the incidence of hypoglycemia was for me to increase the frequency of self-administered blood-sugar tests. It took me far longer than it should have to reach that conclusion, but I should have realized much faster that an alternative treatment would help too. In the mid-1990s, I finally switched from insulin injections to the insulin pump, thanks to my wife who recommended I visit a new endocrinologist. He suggested I switch to the pump for reasons of convenience. However, the switch also had a favorable influence on my research productivity and reduced the total cost of treatment.

The frustrating aspect is that I should have been able to discover my preference for this alternative treatment a lot sooner myself. The argument is that for chronic diseases such as diabetes and arthritis, patients must take a proactive role in disease management. This requires that patients become informed about the disease and about the characteristics of alternative treatments. Patients should be almost as informed as physicians. However, the health care system is still based on the traditional treatment of acute diseases. That is, a patient visits a physician when he or she experiences a problem, the physician diagnoses the problem, and the patient follows the recommended treatment. The patient is mostly passive. After the initial diagnosis of a chronic disease, the patient should take charge and be an active participant in the treatment decisions. Because patient preferences change over time and the available treatment options change as well, the patient should have the opportunity to consider the trade-offs between conflicting characteristics and to learn which treatment is the best fit, subject to medical constraints. Patients should be able to obtain information and to consider trade-offs from a kiosk that sits in the waiting room or on the Web.

By repeating a trade-off exercise periodically, the patient has a chance to discover whether changes in preferred treatment will emerge on the basis of, for example, changes in weights for relevant dimensions (as would have applied to me in the treatment of diabetes). With a continuously updated database with all relevant characteristics for alternative treatments, including new ones, patients also have the opportunity to take advantage of the latest innovations. Importantly, physicians do not have the facility to detect changes in preferences. As a result, most changes occur when patients switch physicians. Indeed, the reason I switched to the insulin pump was the new endocrinologist's argument that I would enjoy the convenience.

I am now eager to promote the idea that patients with chronic diseases, for which there are treatment alternatives that differ on multiple dimensions, have the opportunity to consider the attractiveness of characteristics on a regular basis. On the basis of each individual patient's estimated preference function for treatment characteristics relevant to a given disease, the patient and the physician can learn which treatment option provides the best fit at a given time. Greater patient involvement should result in greater patient satisfaction and enhanced compliance so that health outcomes will improve and total costs might decline. My research with Liana Fraenkel at the Yale School of Medicine is concentrated in this area (Fraenkel, Bogardus, and Wittink 2001, 2003; Fraenkel et al. 2004a, b). We are proposing to conduct experiments to demonstrate the benefits of active patient involvement. The idea of proactive involvement by patients is akin to customers employing decision aids for the selection of the best product or service when there are multiple options that differ on many dimensions.

It is well known that faculty members tend to change the relative emphasis on research, teaching, and other activities over time. Changes in activities occur because of personal preferences but also because of opportunities and institutional demands. An activity that has been fairly constant over time for me is my involvement with doctoral students (Imran Currim and Lakshman Krishnamurthi at Stanford; Naufel Vilcassim, Eric Wruck, Padmini Desikachar, Rishin Roy, John Walsh, Sachin Gupta, Alok Prasad, Rahul Guha, Anil Kaul, Seethu Seetharaman, Niren Sirohi, Sumas Wongsunopparat, and Sev Keil at Cornell; Eijte Foekens, Marco Vriens, Harald van Heerde, Csilla Horvath, and Albert van Dijk at Groningen; and Tom Steenburgh at Yale).

It is not surprising that the Cornell period has the largest number of students whose theses I supervised (the ones in Holland were joint with Peter Leeflang or Michel Wedel). This period was also enormously important personally. Ithaca is a family-friendly place that is conducive to having a happy marriage and happy offspring. It is very important to me that my wife and children realize their own personal and professional goals. The dominant professional themes in my family are health care and education. My wife, originally trained as a nurse, has worked as a childbirth educator and massage therapist, is an advocate of alternative medicines, has an extensive background in Oriental healing arts, and is a dedicated yoga teacher. Our daughter is a licensed doctor who is currently a research fellow focusing on geriatric depression at the University of Pennsylvania's School of Medicine. Our son works at the Center for Responsive Law in Washington, D.C., where one of his projects focuses on legislation related to medical malpractice. Both kids are Cornell graduates, as are their spouses. The diverse health-related professional activities of my family members happen to fit my own newly found predilection. I recently accepted the codirectorship of a proposed Center for Health Management and Policy at Yale University.

Each stage in my professional formation was based on choices I made. However, there were many events over which I had little control that have had a major effect on my career. For example, I never intended to become an academic researcher until a faculty member in statistics at the University of Oregon suggested doctoral study. And the move to the United States was prompted by a scholarship for study abroad that I initially brushed aside. Thus, it seemed appropriate to me to title this essay "An Accidental Venture into Academics." Without the move to the United States, I might not have chosen marketing as a field of study. My interest in applied statistics and econometrics is largely due to accidental exposure to James Reinmuth and Frank Bass.

My accidental venture into academics has been and continues to be a terrific experience. Although I have wondered how I might have fared in business, I cannot imagine that an alternative scenario would have generated the same satisfaction. Of course, if I had pursued a degree in economics in the Netherlands, I might still have ended up doing academic research.

Even though I may be in the mature stage of my life, I am not yet halfway through it. New activities such as the editorship of JMR present great challenges. It is my intention to continue to conduct research and to teach for a long time to come. However, I have learned to welcome seemingly accidental events that occur along the way, because they often lead to interesting developments. I hope my reflections show that having an open mind and being flexible can be very rewarding. Or, if preferred, I believe that I have been very lucky.

( n1) Higher education in the Netherlands has undergone major reorganizations during the past two decades. For example, undergraduate programs now exist, and students have the option to apply for graduate study after they receive a bachelor's degree.

Bass, Frank M. and Dick R. Wittink (1975), "Pooling Issues and Methods in Regression Analysis with Examples in Marketing Research," Journal of Marketing Research, 12 (November), 414-25.

BusinessWeek (1998), "The Best B-Schools," (October 19), 92.

Cattin, Philippe and Dick R. Wittink (1982), "Commercial Use of Conjoint Analysis: A Survey," Journal of Marketing, 46 (Summer), 44-53.

Christen, Markus, Sachin Gupta, John C. Porter, Richard Staelin, and Dick R. Wittink (1997), "Using Market-Level Data to Understand Nonlinear Promotion Effects," Journal of Marketing Research, 34 (August), 322-34.

Currim, Imran S., Charles B. Weinberg, and Dick R. Wittink (1981), "The Design of Subscription Programs for a Performing Arts Series," Journal of Consumer Research, 8 (June), 67-75.

Foekens, Eijte W., Peter S.H. Leeflang, and Dick R. Wittink (1999), "Varying Parameter Models to Accommodate Dynamic Promotion Effects," Journal of Econometrics, 89 (March-April), 249-68.

Fraenkel, Liana, Sidney Bogardus, and Dick R. Wittink (2001), "Understanding Patient Preferences for the Treatment of Lupus Nephritis with Adaptive Conjoint Analysis," Medical Care, 39 (November), 1203-1216.

-----, -----, and ----- (2003), "Risk-Attitude and Patient Treatment Preferences," Lupus, 12, 370-76.

-----, Dick R. Wittink, John Concato, and Terri Fried (2004a), "Are Preferences for Cox-2 Inhibitors Influenced by the Certainty Effect," Journal of Rheumatology, 31 (3), 591-93.

-----, -----, -----, and ----- (2004b), "Informed Choice and the Widespread Use of Anti-Inflammatory Drugs," Arthritis and Rheumatism, forthcoming.

Gomez, Miguel I., Edward W. McLaughlin, and Dick R. Wittink (2004), "Customer Satisfaction and Retail Sales Performance: An Empirical Investigation," Journal of Retailing, forthcoming.

Gupta, Sachin, Pradeep K. Chintagunta, Anil Kaul, and Dick R. Wittink (1996), "Do Household Scanner Panel Data Provide Representative Inferences from Brand Choices?" Journal of Marketing Research, 33 (November), 383-98.

-----, -----, and Dick R. Wittink (1997), "Household Heterogeneity and State Dependence in a Model of Purchase Strings: Empirical Results and Managerial Implications," International Journal of Research in Marketing, 14 (October), 342-58.

Gupta, Sunil (1988), "Impact of Sales Promotions on When, What, and How Much to Buy," Journal of Marketing Research, 25 (November), 342-55.

Huber, Joel C., Dick R. Wittink, John A. Fiedler, and Richard L. Miller (1993), "The Effectiveness of Alternative Preference Elicitation Procedures in Predicting Choice," Journal of Marketing Research, 30 (February), 105-114.

Johnson, Richard M. (1974), "Trade-Off Analysis of Consumer Values," Journal of Marketing Research, 11 (May), 121-27.

----- (1987), "Adaptive Conjoint Analysis," in Sawtooth Software Conference Proceedings. Sequim, WA: Sawtooth Software, 253-65.

Kaul, Anil and Dick R. Wittink (1995), "Empirical Generalizations About the Impact of Advertising on Price and Price Sensitivity," Marketing Science, 14 (Summer), G151-G160.

Keil, Sev K., David Reibstein, and Dick R. Wittink (2001), "The Impact of Business Objectives and the Time of Horizon of Performance Evaluation on Pricing Behavior," International Journal of Research in Marketing, 18 (June), 67-82.

Leeflang, Peter S.H. and Dick R. Wittink (1992), "Diagnosing Competitive Reactions Using (Aggregate) Scanner Data," International Journal of Research in Marketing, 9 (March), 39-57.

----- and ----- (1996), "Competitive Reaction Versus Consumer Response: Do Managers Overreact?" International Journal of Research in Marketing, 13 (April), 103-119.

----- and ----- (2001), "Explaining Competitive Reaction Effects," International Journal of Research in Marketing, 18 (June), 119-38.

-----, -----, Michel Wedel, and Philippe Naert (2001), Building Models for Marketing Decisions. Boston: Kluwer Academic Publishers.

Little, John D.C. (1970), "Models and Managers: The Concept of a Decision Calculus," Management Science, 16 (April), B466-B485.

Lucas, R.E. (1976), "Econometric Policy Evaluation: A Critique," in The Phillips Curve and Labor Markets, K. Brunner and A.H. Meltzer, eds. Amsterdam: North-Holland, 19-46.

Montgomery, David B. and Dick R. Wittink (1980), "The Predictive Validity of Conjoint Analysis for Alternative Aggregation Schemes," in Market Measurement and Analysis, David B. Montgomery and Dick R. Wittink, eds. Cambridge, MA: Marketing Science Institute, 298-309.

Morrison, Donald G. and Jorge Silva-Rossa (1995), "A Latent Look at Empirical Generalizations," Marketing Science 14 (3, Part 2), G61-G70.

Naert, Philippe and Peter S.H. Leeflang (1978), Building Implementable Marketing Models. Leiden: Martinus Nijhoff. Orme, Bryan K. (1998), Reducing the Number-of-Attribute-Levels Effect in ACA with Optimal Weighting. Sequim, WA: Sawtooth Software.

Reinmuth, James E. and Dick R. Wittink (1974), "Recursive Models for Forecasting Seasonal Processes," Journal of Financial and Quantitative Analysis, 9 (September), 659-84.

Sirohi, Niren, Edward M. McLaughlin, and Dick R. Wittink (1998), "A Model of Consumer Perceptions and Intentions for a Supermarket Retailer," Journal of Retailing, 74 (Summer), 223-45.

Srinivasan, V., Dick R. Wittink, and Bruce M. Zweig (1981), "Decision Aids for MBA Program Admissions: Predicting Academic Performance," Research Paper No. 610, Graduate School of Business, Stanford University.

Steenkamp, Jan-Benedict E.M. and Dick R. Wittink (1994), "The Metric Quality of Full Profile Judgments and the Number-of-Levels Effect in Conjoint Analysis," International Journal of Research in Marketing, 11 (June), 275-86.

Van Heerde, Harald J., Sachin Gupta, and Dick R. Wittink (2003), "Is 75% of the Sales Promotion Bump Due to Brand Switching? No, Only 33% Is," Journal of Marketing Research, 40 (November), 481-91.

-----, Peter S.H. Leeflang, and Dick R. Wittink (2000), "The Estimation of Pre-and Post promotion Dips with Store-Level Scanner Data," Journal of Marketing Research, 37 (August), 383-95.

-----, -----, and ----- (2001), "Semiparametric Analysis of the Deal Effect Curve," Journal of Marketing Research, 38 (May), 197-215.

-----, -----, and ----- (2002), "How Promotions Work: Evolutionary Model Building of SCAN*PRO," Schmalenbach Business Review, 54 (July), 198-220.

-----, -----, and ----- (2005), "Decomposing the Sales Promotion Bump with Store Data," Marketing Science, forthcoming.

Verlegh, Peeter W., Hendrik N.J. Schifferstein, and Dick R. Wittink (2002), "Range and Numbers-of-Levels Effects in Derived and Stated Measures of Attribute Importance," Marketing Letters, 13 (1), 41-52.

Wittink, Dick R. (1977), "Exploring Territorial Differences in the Relationship Between Marketing Variables," Journal of Marketing Research, 14 (May), 145-55.

----- (1991), "Attribute Level Effects in Conjoint Results: The Problem and Possible Solutions," in First Annual Advanced Research Techniques Forum Proceedings, William D. Neal, ed. Chicago: American Marketing Association, 44-55.

----- (2001), "Market Measurement and Analysis: The First Marketing Science Conference," Marketing Science, 20 (Fall), 349-56.

-----, Michael Addona, William Hawkes, and John C. Porter (1988), "SCAN*PRO: The Estimation, Validation, and Use of Promotional Effects Based on Scanner Data," working paper, Johnson Graduate School of Management, Cornell University.

----- and Philippe Cattin (1981), "Alternative Estimation Methods for Conjoint Analysis: A Monte Carlo Study," Journal of Marketing Research, 18 (February), 101-106.

-----, Joel C. Huber, John A. Fiedler, and Richard L. Miller (1992), "Attribute Level Effects in Conjoint Revisited: ACA Versus Full Profile," in Second Annual Advanced Research Techniques Forum Proceedings," Rene Mora, ed. Chicago: American Marketing Association, 52-61.

----- and Sev K. Keil (2000), "Continuous Conjoint Analysis," in Conjoint Measurement: Methods and Applications, Anders Gustaffson, Andreas Herrmann, and Frank Huber, eds. Berlin: Springer-Verlag, 411-34.

-----, Lakshman Krishnamurthi, and Julia B. Nutter (1982), "Comparing Derived Importance Weights Across Attributes," Journal of Consumer Research, 8 (March), 471-74.

-----, -----, and David J. Reibstein (1990), "The Effect of Differences in the Number of Attribute Levels on Conjoint Results," Marketing Letters, 1 (2), 113-23.

-----, David J. Reibstein, William Boulding, John E.G. Bateson, and John W. Walsh (1989), "Conjoint Reliability Measures," Marketing Science, 8 (Fall), 371-74.

~~~~~~~~

By Dick R. Wittink and Terry Clark, Editor, Southern Illinois University

Dick R. Wittink is George Rogers Clark Professor of Management and Marketing, School of Management, Yale University (e-mail: dick.wittink@yale.edu), and Honorary Professor of Marketing and Marketing Research, Faculty of Economics, University of Groningen, the Netherlands. He thanks many people who made valuable contributions, in particular Marian Wittink, Marsha Wittink, and Mark Wittink.

An Autobiographical Essay: When I Stop Learning, I Will Leave 

Writing an autobiographical essay is both welcome and daunting. In putting these thoughts to paper, I reflect on the path taken and speculate on where it might yet lead me. I suspect that my experience is unique, and I am not sure that my experience will be for everyone to follow. Nevertheless, my hope is that in describing what worked for me, others may be stimulated in some way to find what works for them. As I reflect back on my career, I have no regrets and am happy with how it progressed--it has certainly been interesting, challenging, and rewarding.

In reflecting on artist Christo's sculpture "wrapping" of the Pont Neuf Bridge in Paris, the director of the De Cordova Art Museum in Lincoln, Mass., stated that Christo's work was an "art event" and not a real "sculpture" because it was only up for a few weeks. I met Christo shortly afterwards and asked him about this comment. He reaction was "nonsense," the work was "real sculpture," and the only reason it was up for such a short time was that the insurance for the installation cost him $200,000 a week. This indicates either that artists do not understand their own work or that their critics do not understand it. As a scholar, I may not be able to judge my own work. In this essay, I describe how I view my research. In doing so, I assume that I understand my own work. I hope I am not too close to it to miss the truth.

I begin by sketching out the area of my primary research--new product development. In doing this, I necessarily and incidentally touch on many of the events that influenced my development and career choices and the attributes of the research style that, in my view, have been critical to the success of my work. I close with a description of my current research plans.

I grew up in Wisconsin--the son of a man who believed that "work is a virtue." Working in Dad's business (Urban Steel Buildings) during the summers had a lasting effect on me. Working with the construction crew was hard, and the hours were long. Dad also let me try my hand in the office as a draftsman, cost estimator, and salesman. It was here that I had my first exposure to real marketing. Among other things, I learned how to sell by providing customer solutions (e.g., a roof guaranteed for 20 years against leaks) and by demonstration (e.g., a special pull rivet that forced aluminum roofing panels tightly together all the time). The product needed to reflect superior engineering and have a benefit for the consumer. We positioned our buildings as the highest quality product but also segmented the market with a "value-priced" line of steel buildings to counter low-cost competition. My intention was to go into Dad's business. I believed that engineering and marketing were the critical skills I needed, so I went to University of Wisconsin for a bachelor's degree in mechanical engineering and an MBA in marketing. I graduated in 1964. But the best laid plans sometimes do not work out.

In that same year, I enrolled in the marketing Ph.D. program at Northwestern. Phil Kotler was my advisor. After a year of core work, I needed to begin thinking about a thesis, and Phil got me involved with new product forecasting at Union Carbide Corporation. I began modeling the interaction effects of a new product on existing products and applied a simulation model to a new polyvinyl fluoride product Union Carbide was introducing. Thanks to Phil and a Union Carbide product manager, things fell into place, so that in the second year, I had a draft of my thesis. My wife and I typed it up (on mimeograph masters) and delivered it to the committee before strapping the skis on the car and heading for Aspen, Colo. On returning, we expected to spend another year on the thesis. (One committee member had asked me, "Are you going to turn your thesis in like this?" Fortunately, he was talking about the format of the tables, not the content.) But the thesis was accepted with minor revisions, and we began a quick and late job search.

Fortunately for us, Massachusetts Institute of Technology (MIT) was also late in hiring in 1966, and I was able to convince the hiring committee that there was a good fit between my management science and marketing interests and the MIT philosophy. I was warned by faculty that though MIT was not an easy place to earn tenure, "it was a good place to be from." My response has been my touchstone ever since: "Fine, when I stop learning, I will leave. Perhaps they will ask me to leave before that happens, but it is worth a try." Given the MIT opportunity, I did not go into Dad's business. Dad always said he overeducated me by funding me through graduate school. In any case, I never did stop learning, and MIT did not ask me to leave, so here I am still. After 36 years, I must say that MIT has been a wonderful environment for my entrepreneurial style of research.

The past 35 years have been an exciting time for marketing science and for modeling new product decision support. The challenges of new product design, forecasting, risk management, and launch strategy have fostered a large set of creative and useful models. No need to summarize this literature here, but interested readers will find Design and Marketing of New Products (Urban and Hauser 1993) and other summaries (e.g., Green, Krieger, and Wind 2001; Ulrich and Eppinger 1995) useful.

My new product research proper began at the University of Wisconsin with my MBA thesis "Product Planning in the Aerospace Industry," in which I described the new product processes and generalized a multistage decision sequence for the industry (see Urban 1964). I was fortunate enough to work directly with General Motors'(GM's) and 3M's aerospace divisions. My doctoral thesis at Northwestern on industrial product life cycle forecasting was a modeling effort aimed at understanding interdependencies between new and existing products (see Urban 1966). Drawing on Monte Carlo simulation and chance-constrained programming techniques, I modeled the product line substitution and complementary effects Union Carbide faced in launching and pricing a new chemical product (Urban 1968). This thesis, along with a co-authored textbook that gave a state-of-the-art view of management science in marketing in the late 1960s (Montgomery and Urban 1969) set the stage for my 35 years of research.

After arriving at MIT, my initial research focused on the launch and test market phases but then began migrating to earlier stages, including premarket testing and design and opportunity identification. This development prompted my friend and colleague, Al Silk (Professor at MIT and subsequently Harvard Business School), to quip, "Urban's research has been going backwards for many years." Although this is true, I retain a research interest in all phases of the development process. Figure 1 positions my research efforts in the new product development decision process. Here, I give a stream-of-consciousness description of my efforts and then identify the critical issues in my approach to research.

Test Marketing

In the late 1960s, major new theoretical approaches were being developed in the field of stochastic models. Growing out of a contact with a summer session student from the Noxell Corporation (which sells Noxema and Cover Girl skin care products), I learned that forecasting national sales levels based on test market results, planning the best marketing mix for launch, and tracking test market and launch for diagnosis and control were important problems that were not being adequately addressed. This led to a sponsored research project at MIT and the development of a macro flow model methodology that combined elements of stochastic models, response functions, and empirical data in a managerial tool called "SPRINTER" for managing the new product test market and launch (Urban 1970). I was interested in discovering whether this model would really help managers, so I joined John Little and Len Lodish in a firm called Management Decision Systems Inc. Applications at Noxell and Nabisco helped refine the SPRINTER model and provided great case material for my articles and teaching.

During 1970, I spent a term visiting the Indian Institute of Management in Calcutta and became interested in the management of family planning. After returning to the United States, I began working with the Atlanta Area Family Planning Program and the Centers for Disease Control in an effort to improve the efficiency and effectiveness of their programs. My approach was to elaborate and extend the SPRINTER model and apply it to the trial and adoption of family planning (Urban 1974a). At that time, I had two master's students (Ron O'Connor and Joel Lamstein) who were interested in implementing new management techniques in the public arena, so we formed a nonprofit firm called Management Science for Health. Although I am no longer involved with the company (or its spin-off, John Snow Inc.), it is gratifying to report that they now employ more than 600 people working to improve public health management worldwide.

Premarket Testing

In 1972, Cal Hodock, then market research director at Gillette, called me with an invitation to join him for lunch at Loch Ober, a premier restaurant in Boston. I was somewhat surprised because usually I bought lunch for him, in hopes of garnering MIT-sponsored research funds from Gillette. During lunch, he told me Gillette was looking for a modeling and measurement system to forecast sales of a new product in test market, based on the pretest market availability of the product, packaging, and advertising. He wanted the research to cost (on an ongoing basis) less than $25,000 and the forecast to be delivered three months after the project started. On the basis of the complexities I had seen in test market tracking and forecasting, I told him it was impossible. He was persistent. In the end, he persuaded Al Silk and me to look at his problem by dangling $40,000 for sponsored research funds at MIT. How could we refuse? Gillette's need, combined with the emerging logit modeling technology at MIT (McFadden 1970), led us to develop a convergent premarket forecasting approach based on measured changes by two independent models: ( 1) preference change and ( 2) laboratory-simulated trial and repeat purchasing. The result was the ASSESSOR model for forecasting the sales of new packaged goods (Silk and Urban 1978), and over time its validation was based on Management Decision Systems applications (Urban and Katz 1983).

In parallel with the validation work on consumer packaged goods, I was pursuing the application of pretest market forecasting to consumer durables. This grew out of discussions with a student of mine (John Dables) who worked at the Buick division of GM. He told me that the risks involved in developing a new automobile product were much greater than those involved with developing a packaged goods new product, because the investment was so much larger. The whole picture was further complicated by the lack of test marketing for automobiles. The thought occurred to me, why not apply the ASSESSOR methodology to consumer durables? Our discussion led to a five-year Buick-sponsored research project at MIT, which resulted in a durables ASSESSOR model and applications to premarket automobile forecasting based on an early production line version of the new automobile (Roberts and Urban 1988; Urban, Hauser, and Roberts 1990; Urban, Hulland, and Weinberg 1993). Good forecasting results were achieved, but top managers at GM argued that though our analysis could improve the launch, the forecast came too late in the process, because once the car existed in initial production line versions, the launch commitment was virtually assured. The costs were sunk, and on a marginal basis it was almost always profitable to go forward. This was forcefully brought home to me when we predicted in 1986 that the new down-sized Buick Riviera sales would be half of the old level rather than the hoped-for doubling of sales. Buick introduced the car anyway. Although we were glad to have the opportunity for validation (sales dropped to .4 of the old level), we were indeed too late in the process to stop the program.

In the late 1980s, Hyper Card was developed at Apple, and MIT's Media Lab had invented the basic elements of surrogate travel. In 1990, we began an effort to use interactive multimedia to create a virtual automobile market of the future, before the new car was built. We put the customer in the future environment with full information and ability to control the search, and we measured responses to predict future sales before the production commitment was made. I called this "information acceleration" (IA). The first application was to electric vehicles at GM (Urban, Weinberg, and Hauser 1996), and on the basis of application and validation experience (Urban et al. 1997), the potential of this model and measurement methodology was encouraging. Initial field testing of the GM EV-1 two-seat sports car was done at MIT with the IA model. The final forecasting was done during 1992-93 through a consulting firm called Marketing Insight Technology Inc., which I started with a few of my students to implement IA concepts. The forecasts, based on expanded proprietary surveys, were for low sales in the 1998-2000 period (fewer than 1000 units per year) and indicated that the real demand was for an economy car with a hybrid power system for reliability (an electric motor plus a small gas-powered generator). General Motors did not go into production with this vehicle but rather custom built units for sale in California and selected other locations. In 1999, the sales of the EV by GM were about 700 units, surprisingly close to the forecast, given the vagaries of this market. I was given permission to publish these results, so the results considerably strengthened the MIT initial research for publication. In 1999, I was at an American Marketing Association conference in San Diego and saw the Toyota Prius (a four-door hybrid electric vehicle) in the hallway and a sign announcing a presentation of the Prius development story. I was pleased to see that the Toyota product manager had used the IA, on the basis of my publications only, to forecast the sales of the Prius hybrid economy car and had found a real market opportunity.

Design

As the seminal work on perceptual mapping was appearing in the 1970s, there seemed to be a natural fit to new product design. The notion of a "core benefit proposition" could be represented in the positioning and in a model called PERCEPTOR. I made an early attempt to link positioning to new product sales potential and extended this model for marketing of the MIT health maintenance organization (Hauser and Urban 1977; Urban 1975). I continued research on product design in an effort to integrate Von Hippel's lead user notions with market research methods (Urban and Von Hippel 1988) and apply it to industrial product (i.e., CAD/CAM systems for electronic printed circuit boards at Computer Graphics Inc.) innovation and diffusion from lead users to other customers. This theme has carried through to current research, in which I am putting lead users on an Internet design pallet to configure their ideal pickup trucks.

Opportunity Identification

Through the 1980s, I became convinced that marketers needed not only tools to help effectively forecast and design products but also tools to help identify strategic opportunities. My first project in this area focused on market definition. This returned me to my original interest in product lines and interdependency. I tried to define a hierarchical market structure that created segments in which intra-segment competition existed but intersegment competition was limited, so that little customer switching among segments occurred. This system was called PRODEGY and addressed PRODuct stratEGY by examining the coverage and duplication of a product line (Urban, Johnson, and Hauser 1984).

The second project grew out of the empirical experience I gained from applications of ASSESSOR. Contrary to the predictions of perceptual mapping models, I noticed that second brands in a market rarely received the same share as successful first entrants, even if they had parity positioning and allocated similar resource levels to advertising and promotion. This led to a statistical cross-sectional analysis of the effects of order of entry on market share (Urban et al. 1986). My coworkers and I confirmed this order-of-entry effect in a time-series cross-sectional analysis of test market scanner data (Kalyanaram and Urban 1992) and ethical pharmaceuticals (Berndt et al. 1995). Our work in this area was contemporaneous with the PIMS data analysis and led to interesting insights as the results were integrated into the growing literature on order of entry (Kalyanaram, Robinson, and Urban 1995).

Launch and Life Cycle Management

Recently, I returned to the topic I had begun with while at Northwestern--the life cycle phase. I have developed a set of concepts for trust-based marketing over the life cycle (Urban, Sultan, and Qualls 2000), in which the use of a virtual advisor on the Internet provides customers with full and accurate information and unbiased advice in a private, secure, branded, friendly, and easy-to-use system. This project grew out of a realization that my IA ideas could be used to help customers make better decisions on existing products, as well as test new products. Inverting IA gives a system that, when supplemented by a personal advisor, provides a trust-based marketing tool. Vince Barabba, Director of Market Planning at GM, encouraged me in this work and funded an MIT-sponsored research project. After developing a prototype, we tested it in the field by application with 300 customers to pickup truck purchasing. Initial results indicated substantial increases in trust, and presumably sales can be earned through the Internet virtual advisor. General Motors is now considering implementing such an advisor on one of its Internet sites.

This chronology of my research indicates several threads I believe have been important in the success of my work. I briefly discuss some of these.

Managerial Need Input and Implementation

My research style is inductive, so I found it natural to work closely with managers making real decisions. I have always been impressed with the knowledge and insight managers accrue in facing tough decisions. As a marketer, I instinctively thought in terms of "customer needs" as I defined my customers for the analytic models I worked on as managers and tried to involve them early in the design of decision support models. In 1980, while I was reporting the results of a second PERCEPTOR study at Dow Corporation, the group product manager leaned over and said, "Tell us something we do not know this time." It was new to me, but old stuff to him. Coping with implementation problems gave me a growing awareness of decision needs, so my following projects could be better fitted to the changing managerial decision requirements. Building models and applying them should be considered an organizational change process, not an exercise in mathematical gymnastics (Urban 1974b; Urban and Karash 1971). As a result, implementation should be considered from the start of the project to beyond its academic completion if academics are to keep their research relevant and improve the practice of marketing.

Equally important, this manager orientation can help generate funds for research assistants, computers, software, and large databases. I have also found that real applications after publication are useful in assuring that models are used and that evolutionary model extensions can create a positive benefit-cost ratio for managers.

Sometimes I was involved with applications as a consultant, but more often I have worked within companies I have founded with my students. We founded these companies to implement the new technologies, and though it is nice to be economically successful, my real motivation was to change the practice of management. For example, ASSESSOR was implemented by Management Decision Systems (and subsequently by Information Resources Inc., which bought Management Decision Systems, and then by MARC, which bought the ASSESSOR business from Information Resources Inc.). But the Journal of Marketing Research publication (Silk and Urban 1978) on ASSESSOR was used by Research International Inc. and Novaction Inc. to build competitive services. My best estimate is that more than 3000 consumer packaged goods products have been tested by the ASSESSOR methodology and its derivatives. I doubt that ASSESSOR would ever have been applied without Management Decision Systems as a proving ground. Similarly, IA was implemented by Marketing Technology Inter-face, as well as by others (e.g., Toyota, Intel, various market research companies), on the basis of the Journal of Marketing Research publications (Urban et al. 1996; Urban, Weinberg, and Hauser 1997). My role in these companies was limited to less than one day per week, but this was enough to help design the implementation procedures, interact with clients on design issues, and identify new research opportunities. Overall, my intent has been to build new decision tools that reflect customer needs and result in better products and reduced risk in new product innovation.

Matching Needs to Theory

Interacting with managers has been important to me to understand their needs. However, successful research also requires matching these needs to emerging theories and methods. When the two are in sync, the ensuing research can advance the state-of-the-art of marketing science as well as affect practice. Whether it is logit modeling, multidimensional scaling, utility theory, artificial intelligence, or virtual reality, I have always looked for problems that lend themselves to analysis through the most recent management and behavioral science technologies. I view this matching of theory to practice as a creative process.

Not all problems are interesting academically. Pure application projects may be useful but lie in the consulting domain. Pure theory can be important work, but I have tended to examine problems that require application of the latest theory. I think this tendency reflects my engineering training. However, I have often found that modeling requires both theory extension and innovation in estimation. For example, ASSESSOR came from Gillette's managerial problem, but the solution was in the then-new logit analysis. When logit was first applied, McFadden (1970) was just developing the maximum likelihood algorithms. I believe that ASSESSOR was the first application in the marketing of logit analysis. This theory enabled us to develop a new convergent forecasting methodology for premarket analysis. Similarly, when modeling the problem of premarket forecasting of new automobile sales, multiattribute utility theory was used and extended as a modeling framework (Roberts and Urban 1988). My recent work to develop trust on the Internet uses artificial intelligence theory to build a trusted advisor for automobile purchasing (Urban, Sultan, and Qualls 2000).

Power of Empirical Data

I have been a heavy user of measurement and empirical data. Whether it be test market, laboratory simulation, survey, market experiments, or virtual reality data, I have felt compelled to measure customer response. I have also been diligent in testing my model predictions. This is a difficult validation process but a critical one if marketing science is to progress. Often, these empirical efforts require creativity and innovation in measurement methodology and persistence in obtaining response and validation data, but the research power gained is well worth the effort.

Research Risks

A sense of adventure, entrepreneurship, and intellectual flexibility has served me well in my efforts to match theory with managerial relevance. I have generally avoided small epsilon extensions of existing work in favor of major problems that have not been extensively studied. This is ambitious and risky, because such innovative work takes a long time--especially if it is empirically based. Reviewers may not always understand the value of the new work, or else they find many methodological problems that can only be resolved by further research. Articles may be rejected or may need to be revised and aggressively defended. But some of them may win prizes. As I see it, the low-risk way to publication is to extend previous work in a field. When this is the case, previous contributors are usually (and naturally) chosen as reviewers, and they find it easier to understand and accept the extension, even if it is not a major breakthrough. Certainly, this evolutionary research path has moved the field forward. However, it is not my style. I like to find the big problems managers face and see if I can solve them. In this effort, I have often found that the literature constrains my thinking. As a result, I often work on a problem for months before reading the existing literature and modify my efforts to profit from prior research. It is a balance between creativity and constraint. Existing theory should be used, but for me the drivers are the problem and creativity rather than the placement the new work in the structure of the existing publication stream.

Programmatic Research

I am a research planner. My method is to lay out my research intentions over one-and five-year time horizons and examine how they fit into the accomplishment of my overall research goal--improving the productivity of new product development and advancing the art of marketing science. This sometimes calls for long-term projects--most of my models have a five-year or longer development time frame. (For example, ASSESSOR was started in 1973 and published in 1978, IA was started in 1990 and published in 1996 and 1997, and trust-based marketing was started in 1996 and published in 2000.) This may not maximize the number of publications, but those that do come out the end of the pipeline can be significant. Fortunately, MIT has been patient and has tenure criteria that do not depend solely on the amount of publication.

Balance

Real academic success requires that we balance research, teaching, and administration. My annual plans have always incorporated activities from each of these areas. Most often, for me, the one that slips is research. Academics need to be disciplined to keep research priority high. The demand for high-quality teaching has continually risen over time. To help cope with this, I have tried to tie my teaching to my research to make the most efficient use of my time. For example, Dave Montgomery and I taught a marketing modeling course based on our book Management Science in Marketing (1969). In many ways, that course was the foundation for my future modeling. I taught the first new product marketing and development course at MIT. This was synergistic with my ASSESSOR research and led to the development of teaching materials such as a textbook with John Hauser titled Design and Marketing of New Products (1993). When my focus became more strategic, I found that teaching marketing strategy and co-authoring a text and case book with Steve Star, Advanced Marketing Strategy (Urban and Star 1991), helped me understand and teach the wider context of marketing and analytical support. The secret for me is balancing activities and building synergy.

In administration, I served as deputy dean (1987-91) and then dean at MIT Sloan (1993-98). These were the most difficult times for me to achieve balance. I found as dean that I could continue research work (e.g., IA, trusted advisors on the Internet) I started before taking on the dean's responsibilities, but it was difficult to start new projects. The compensating benefit was that I learned a lot about operational management, gained a wider understanding of other fields of management, and was exposed to a wide array of interesting academics and top managers. It was clear in 1998, however, that I needed to choose between becoming a full-time administrator and a researcher/teacher. I chose the latter, but I am confident that the experience in the dean's office will make me a better teacher and give me a wider perspective on research. A sabbatical between each of the two administrative positions was critical to rebalancing and energizing my research activity.

Another critical balance is between work and personal time allocations. The academic tenure system will push a professor so that it is difficult to have a life outside of work. I have found that it is critical to have a balance among work, family, and personal time. My wife Andrea recognized this problem early and signed me up for a sculpture course in 1970 at the local art museum. I enjoyed the class, and sculpture became a major hobby for me. I do mostly large steel and bronze work, but some stone and wood carving. I have more than 50 pieces in my yard and house. It is rewarding to come home from school to see the tangible results of cutting and welding steel for a couple hours, rather than the almost invisible results of daily research. Only after a long and frustrating period of research, writing, and revising is an article published. For me, sculpture provides unconstrained and immediate results. In addition, I believe sculpture has helped my research. Building a mathematical model taps the same sort of creative energy I find necessary in abstract sculpture. I have enjoyed sculpture, skiing, and sailing, but in retrospect I have put too much time into my work. If I had it to do over, I would put in fewer hours at MIT and more at home with my family. This would have been easier if the work had not been so exciting.

Great Co-authors

Good co-authors are an intellectual inspiration, and I have had some of the best. I am greatly indebted to them. I must acknowledge John Hauser (who has written more joint articles with me than anyone) for his rigor, scholarly standards, tight writing, and creative input. I have also benefited greatly from my other academic co-authors, including but not limited to Montgomery, Silk, Von Hippel, Star, Robinson, Berndt, Qualls, and Sultan (in chronological order of publication). It is also important to recognize my student co-authors, who have probably received less credit than they deserve for their input (e.g., Weinberg, Kalyanaram, Bohlman, Hulland, Roberts, Carter, Gaskin, Mucha, Johnson, Katz, and Karash, in reverse chronological order). Although I have generally not co-authored with line managers who have contributed to my work, special contributions were made by several of them (e.g., Ed Sellars of Noxell, Cal Hodock of Gillette, Tom Hatch of Miles Labs, John Dables of Buick, Roberta Chicos of MTI, and Sean McNamara and Vince Barabba of GM). Finally, John Little has been my mentor, and even though we have never co-authored a major article, he influenced every one of my works through his example, comments, and criticisms. John is a straight-thinking, rigorous scholar who believes in research paying off in practice.

My current five-year research plan calls for extensions to my new product modeling and a major thrust toward developing models to exploit the potential of the Internet for marketing managers.

My work on trust-based advisors raises a challenge for managers. If trust-based marketing involves fair comparison between alternatives, what does a firm do if its products are not the best available? One solution is to back off of trust and push what you have. A better long-term solution is to find unmet needs and build the highest quality products to fill those needs. In my trust work, we applied the virtual advisor to pickup trucks, so it was natural with GM's research support to extend the research to discover whether we could "listen in" to the dialogue between the advisor and the real customer to find unmet needs. We applied utility theory to identify the most preferred truck and the level of utility. We posited that if the utility of the most preferred truck goes down after a question in the dialogue, this indicated an unmet need. For example, if a consumer wants a small truck, the Mazda might be the most preferred. After the consumer indicates that he or she wants to tow a boat, the Chevrolet Blazer may be the most preferred, but the utility is likely to have gone down. This drop indicates the need for a small truck that can tow. This unmet need is explored by a virtual engineer who asks the customer for details about the need (e.g., How much does the boat weigh? Why do you want a small truck? For gas efficiency? Parking?). This virtual engineer provides detailed input to the platform design team. A final component in this analysis is to put the customers on a design pallet and let them specify any truck they want (e.g., a mid-sized truck that can tow 6000 pounds and still be easily parked). This system has been estimated on the basis of 1000 customer interviews; the results indicate that it can identify significant new opportunities and the need-identification algorithms are robust (Urban and Hauser 2002).

The Internet is a risky area for research, because it is so volatile and we do not have much research banked in this area, but I believe that it will be a major additional channel for marketers in the future. Currently with the support of the Inter Public Group, my colleagues (Fareena Sultan, Venky Shankar, and Iakov Bart) and I are analyzing 6800 consumer evaluations of 24 leading Internet sites to find the determinants of trust on the basis of 120 cue assessments (e.g., security, privacy, personalness, information, navigation, advice, brand). This empirical analysis will supply under-standing to enable effective experimentation of site design and consistency with other communication channels. We have begun experiments on the Internet to test the causal nature of trust-building in a site. This work is funded by Intel, and I hope in the next several years, in collaboration with others in the MIT marketing group (John D.C. Little, John Hauser, and Duncan Simester), to extend it to full adaptive marketing. We plan to draw on reinforcement learning (Sutton and Barto 1998). I also would like to study the use of the Internet as a direct manufacturer sales channel, complementary to the existing distribution system (e.g., Palm sells direct at the same price as through existing retail stores). Finally, I plan to study the implications of increasing customer power and the paradigm shift from push to trust-based marketing that this may precipitate.

The future is full of exciting marketing opportunities, and if we can effectively integrate theory, practice, empirical data, and creativity in research, we can improve the efficiency and effectiveness of marketing.

Figure 1: New Product Process, Selected Works

Berndt, Ernst R., Linda Bui, David R. Reile, and Glen L. Urban (1995), "Information, Marketing, and Pricing in the U.S. Antiulcer Drug Market," AEA Papers and Proceedings, 85 (May), 100-105.

Green, P.E., A.M. Krieger, and Y. Wind (2001), "Think Years of Conjoint Analysis: Reflections and Prospects," Interfaces, 31 (3), 56-73.

Hauser, John R. and Glen L. Urban (1977), "A Normative Methodology for Modeling Consumer Response to Innovation," Operations Research, 25 (July/August), 579-619.

Kalyanaram, G., W.T. Robinson, and Glen L. Urban (1995), "Order of Market Entry: Established Empirical Generalizations, Emerging Empirical Generalizations, and Future Research," Marketing Science, 14 (3), G212-G221.

------ and Glen L. Urban (1992), "Dynamic Effects of the Order of Entry on Market Sham, Trial Penetration, and Repeat Purchases for Frequently Purchased Consumer Goods," Marketing Science, 11 (Summer), 235-50.

McFadden, Daniel (1970), "Conditional Logit Analysis of Qualitative Choice Behavior," in Frontiers in Econometrics, P. Zarembka, ed. New York: Academic Press, 105-42.

Montgomery, David B. and Glen L. Urban (1969), Management Science in Marketing. Englewood Cliffs, NJ: Prentice Hall.

Roberts, J. and Glen L. Urban (1988), "Modeling Multiattribute Utility, Risk, and Belief Dynamics for New Consumer Durable Brand Choice," Management Science, 34 (February), 167-215.

Silk, Alvin J. and Glen L. Urban (1978), "Pretest Market Evaluation of New Packaged Goods: A Model and Measurement Methodology," Journal of Marketing Research, 15 (May), 171-91.

Sutton, R.S. and A.G. Barto (1998), Reinforcement Learning. Cambridge, MA: MIT Press.

Ulrich, K. and S.D. Eppinger (1995), Product Design and Development. New York: McGraw-Hill.

Urban, Glen L. (1964), "Product Planning in the Aerospace Industry," MBA thesis, University of Wisconsin, Madison.

------ (1966), "A Quantitative Model of New Product Planning with Special Emphasis on Product Interdependency," doctoral thesis, Department of Marketing, Northwestern University.

------ (1968), "A New Products Analysis and Decision Model," Management Science, 14 (8), 490-517.

------ (1970), "SPRINTER Mod III: A Model for the Analysis of Frequently Purchased Consumer Goods," Operations Research, 18 (September/October), 805-54.

------ (1974a), "A Model for Managing a Family Planning System," Operations Research, 22 (March/April), 205-34.

------ (1974b), "Building Models for Decision Makers," Inter-faces, 4 (3), 1-12.

------ (1975), "PERCEPTOR: A Model for Product Positioning," Management Science, 21 (8), 849-58.

------, T. Carter, S. Gaskin, and Z. Mucha (1986), "Market Share Rewards to Pioneering Brands: An Empirical Analysis and Strategic Implications," Management Science, 32 (6), 645-59.

------ and John R. Hauser (1993), Design and Marketing of New Products, 2d ed. Englewood Cliffs, NJ: Prentice Hall.

------ and ------ (2002), "Listening In to Find Unmet Customer Needs and Solutions," working paper, Alfred P. Sloan School of Management, Massachusetts Institute of Technology.

------, ------, and J. Roberts (1990), "Prelaunch Forecasting of New Automobiles," Management Science, 36 (4), 401-21.

------, John S. Hulland, and Bruce D. Weinberg (1993), "Pre-market Forecasting for New Consumer Durable Goods: Modeling, Categorization, Elimination, and Consideration Phenomena," Journal of Marketing, 57 (2), 47-63.

------, P.L. Johnson, and John R. Hauser (1984), "Testing Competitive Market Structure: A Methodology Based on Product Deletion Probabilities," Marketing Science, 3 (2), 83-112.

------ and R. Karash (1971), "Evolutionary Model Building," Journal of Marketing Research, 8 (1), 62-66.

------ and Gerald M. Katz (1983), "Pretest Market Models: Validation and Managerial Implications," Journal of Marketing Research, 20 (August), 221-34.

------ and S.H. Star (1991), Advanced Marketing Strategy. Englewood Cliffs, NJ: Prentice Hall.

------, F. Sultan, William Qualls, (2000) "Placing Trust at the Center of Your Internet Strategy," Sloan Management Review, 42 (1), 39-48.

------ and E. Von Hippel (1988), "Lead User Analysis for the Development of New Industrial Products," Management Science, 34 (5), 569-82.

------, Bruce D. Weinberg, and John R. Hauser (1996), "Premarket Forecasting of Really-New Products," Journal of Marketing, 60 (January), 47-60.

------, ------, William J. Qualls, Bruce D. Weinberg, Jonathan D. Bohlmann, and Roberta A. Chicos (1997), "Information Acceleration: Validation and Lessons from the Field," Journal of Marketing Research, 34 (February), 143-53.

~~~~~~~~

By Glen L. Urban, Professor of Management, Sloan School of Management, Massachusetts Institute of Technology. and Terry Clark, Editor. Southern Illinois University

An Empirical Analysis of the Determinants of Retail Margins: The Role of Store-Brand Share 

The authors develop and test a model of the key determinants of margins that retailers earn on national brands and store brands. They particularly focus on the impact of store-brand share on percentage margin, dollar margin per unit, and total dollar margin of the retailer. The authors find not only that percentage retail margins on store brands are higher than on national brands but also that high store-brand share enables retailers to earn higher percentage margins on national brands. However, the dollar margin per unit may be smaller for store brands because of their lower retail price. Furthermore, heavy store-brand users contribute much less to the total dollar profit of the retailer than do light store-brand users. The authors conclude that it is important for retailers to retain a balance between store brands and national br ands to attract and retain the most profitable customers.

During the past decade, retail profit margins in the packaged goods industry have been the subject of substantial discussion and research. On the one hand, much has been written about the increasing power of retailers and their proclivity to negotiate lower wholesale prices and higher trade allowances from manufacturers (e.g., Alpert, Kamins, and Graham 1992; Buzzell, Quelch, and Salmon 1990; Harrison 1999a, b). On the other hand, there is growing evidence that aggregate retail profit margins do not support the purported shift in power: Retail industry margins are either shrinking or remaining steady (Farris and Ailawadi 1992; Messinger and Narasimhan 1995).

A major topic at the center of much of the discussion about retailer power and profitability relative to that of manufacturers is the growth of store brands, for which average U.S. market share was up from 15.3% in 1988 to 20% in 1998 (Dunne and Narasimhan 1999; Marketing News 1995). In a survey of retailers, Cannondale Associates (1996) found that more than 70% of retailers considered store-brand growth an important issue (Wellman 1997). In a Discount Merchandiser (1996) survey, retailers rated "better profit margins" as the most important reason for carrying store brands. No wonder that several U.S. retailers now sell store brands in many product categories. In addition, researchers have provided both theoretical and empirical insights about when store brands should be introduced (e.g., Raju, Sethuraman, and Dhar 1995) and about the determinants of store-brand share across categories and retailers (e.g., Dhar and Hoch 1997; Hoch and Banerji 1993).

Despite all the interest in retail profitability, little is known about either the determinants of retail margins across product categories or the specific relationship between store-brand share and retail margins after a store brand has been introduced. The purpose of this article is to fill the gap in the literature by conducting a systematic empirical examination of the determinants of margins retailers earn on national brands and store brands and of the impact of store-brand share on these margins.

We address three specific clusters of research questions. First, what are the key determinants of retail margins across product categories? How do factors such as manufacturer concentration, advertising, and distribution affect retail margins on national and store brands? When these factors are accounted for, do retailers earn higher percentage margins on national brands in categories in which they have strong store-brand share than in categories in which their store-brand share is weak? Second, although it is well known that retailers earn higher percentage gross margins on store brands than on national brands, how and why does this margin differential vary across product categories? Does the differential hold up when considering percentage contribution margins net of the direct product costs retailers incur and when considering dollar rather than percentage margins? Third, how does store-brand share affect the retailer's total dollar profit? Do consumers who are heavy users of store brands contribute more or less than light users or nonusers of store brands to the total profit of the retailer? To answer these questions, we use sales, share, and profit margin data from two major retail chains that are supplemented with category characteristics from other sources. The article is organized as follows: We provide the motivation for our work by reviewing prior research that is relevant to each of our research questions, and we show why the answers are not obvious. We then describe the data used in our analysis and subsequently address our research questions. We develop and estimate a model, including key determinants such as advertising, distribution, and store-brand share, of the gross and net margins that retailers earn on national brands and store brands. We then compare retail percentage and dollar margins on store brands and national brands across several product categories. We explore the relationship of store-brand share to total retail profit by comparing the average dollar profit that retailers obtain from customers who are nonusers, light users, and heavy users of store brands. We conclude with a discussion of our findings and implications for further research.

Determinants of Percentage Retail Margins

According to basic industrial organization theory, the margin that a market player can earn is a function of the player's market power. In the context of the packaged goods industry, the relative market power of retailers versus manufacturers determines how total channel profit is split between the two and the margins that retailers can earn (e.g., Kadiyali, Chintagunta, and Vilcassim 2000). In turn, the relative market power of retailers versus manufacturers is influenced by several key variables, such as national brand concentration, advertising spending, distribution, and store-brand share, all of which vary across product categories. As a result, retail margins also vary significantly across categories. The literature contains some important conceptual and analytical work on the inverse relationship between manufacturer and retail margins and on the role of advertising and distribution in the relationship (Abela and Farris 1999; Albion and Farris 1981, 1987; Lal and Narasimhan 1996; Steiner 1973, 1978, 1993). In recent years, researchers have also developed structural models of optimal retailer pricing from first principles of profit maximization (e.g., Besanko, Gupta, and Jain 1998; Chintagunta, Bonfrer, and Song 2002; Kadiyali, Chintagunta, and Vilcassim 2000). This stream of research provides a useful and flexible way to describe the type of pricing game played by a retailer and a few national brand manufacturers in a single category. However, we are not aware of any comprehensive empirical analysis of factors explaining cross-category variations in retail margins on national brands and store brands. Such a study would complement Shankar and Bolton's (2003) analysis of the variation in retailer pricing strategies across brands, categories, stores, and markets.

There is much discussion in the theoretical literature about a determinant of retail margins that is of particular interest in our research: store-brand share. In general, researchers agree that introducing a store brand can enable retailers to obtain lower wholesale prices from national brand manufacturers and therefore earn higher margins on national brands. However, there is disagreement about the relationship between store-brand share and the retailer's margin on national brands after the store brand has been introduced. On the one hand, Mills (1995) predicts that the higher the market share of the store brand, the lower is the wholesale price of national brands and therefore the higher is the margin that the retailer can obtain on national brands. On the other hand, Narasimhan and Wilcox (1998) predict a negative relationship between store-brand share and the retailer's margin on national brands. They distinguish between the exercised and unexercised threat of a store brand and argue that the retailer's ability to extract channel profit lies in the market share that the store brand can potentially attain, not in its actual market share. Morton and Zettelmeyer (2000) make a similar argument: They do not expect that store-brand market share is necessarily related to retail margins from national brands.

Despite the divergent theoretical predictions, empirical evidence of the relationship between store-brand share and retail margins is scarce. Narasimhan and Wilcox (1998) provide some empirical results, but their analysis has some important limitations that must be overcome. First, their analysis is based on aggregate data across retailers that are likely to be heterogeneous in many aspects that determine their store-brand success and their bargaining ability with national brands. Second, their analysis does not use actual margin data from specific retailers; instead, it uses estimates of average retail margins from consumer expenditure survey data and does not separate store-brand and national-brand margins. Third, it uses penetration of store brands rather than of market share, but penetration is just one component of share (Ailawadi, Lehmann, and Neslin 2001). Finally, their analysis omits several key determinants of store-brand share and margin. Thus, it is not clear whether high store-brand share increases retail margins on national brands or whether it is better to have a store brand on the shelf as an unexercised threat.

We develop an econometric model of a retailer's margin on national and store brands, which we estimate using data from individual retail chains. Because retailers may put more effort into store brands in categories with higher margins or may deliberately raise their national-brand prices and thus their national-brand margins to increase store-brand share, there may be reverse causality between retail margins and store-brand share. We control for this reverse causality.

The Store Brand--National Brand Margin Differential

In general, it is known that retailers earn higher percentage gross margins on store brands than on national brands (e.g., Handy 1985; Hoch and Banerji 1993); this primarily is because private label suppliers have little market power, in contrast with national brands. Private label suppliers operate in a competitive market with no product differentiation and thus sell to retailers at a price that is close to their marginal cost. In addition, manufacturers' advertising costs are high for national brands, and these costs are reflected in higher wholesale prices of the brands. In addition, a retailer has somewhat of a local monopoly on its store brand because competing retailers do not carry the same store brand, though they do carry the same national brands.

However, some questions about the store brand--national brand margin differential remain unanswered. First, gross and net contribution margins are often not highly correlated in practice (Borin and Farris 1990), yet there is little evidence about the difference in net margins between store brands and national brands when the retailer's direct product costs are taken into account. The single exception that we know of is Corstjens and Lal's (2000) description of a Canadian retailer's beverage category, which suggests that net percentage retail margins on store brands can be lower than on national brands. We therefore compare not only gross margins but also net contribution margins that retailers earn on national and store brands.

Second, the percentage margin differential is certainly not equal across all categories, but we are not aware of any work that examines how and why this differential varies across categories. We therefore use our cross-category determinants of retail margins on national brands and store brands to explain variations in the store brand-national brand margin differential.

Third, attention has been focused on percentage margins, probably because they can be readily compared across categories, but retailers should be concerned at least as much with dollar margins. Even if retail percentage margins are higher for store brands than for national brands, corresponding dollar margins may not be. Retailers pay lower wholesale prices for store brands, but store brands are sold at significantly lower retail prices than national brands are. If the higher percentage margin on store brands does not offset the lower retail price, dollar margin per unit may be lower for store brands than for national brands. This may be particularly true in the face of increasing store-brand competition, not just in the grocery channel but also in the mass-merchandiser and drugstore channels (Johnson 1994; Matthews 1995). Therefore, we also compare dollar margins from store brands and national brands.

The Profitability of Store-Brand Users

Retailers should also be concerned with total dollar margin in the store. How much do heavy store-brand users, compared with light or nonusers, contribute to the total dollar margin of the retailer? Even if the retailer's dollar margin per unit is higher from store brands, the total profitability of a heavy store-brand user depends on how many items and which other items he or she buys from the retailer.

Several researchers have suggested that store brands are associated with better store image and higher store loyalty (Ailawadi, Neslin, and Gedenk 2001; Corstjens and Lal 2000; Steenkamp and Dekimpe 1997), which would have a positive effect on retail sales and total dollar margin because loyal consumers buy a larger share of their total grocery requirements from the store. However, it may also be that heavy users of store brands are loyal to store brands in general, not necessarily to the store brand of a particular retailer. Furthermore, if store brands attract budget-constrained, price-conscious, or deal-prone consumers whose market baskets are small or contain other low-margin/loss-leader products, store-brand users may contribute less than nonusers do to the total dollar profit of the retailer (e.g., Quelch and Harding 1996). Although the demographic and psychographic correlates of store-brand proneness have been studied (Ailawadi, Neslin, and Gedenk 2001; Richardson, Jain, and Dick 1996), we are not aware of any research that examines store-brand users' profitability. We explore this issue by comparing the total profitability of heavy store-brand users with that of light store-brand users and nonusers.

We used data from two anonymous retail chains and conducted all our analyses for each retailer separately. This provided a valuable validation test for our results, especially because the two retailers differ in many respects: One is a grocery retail chain and the other is a drug retail chain; one practices everyday low pricing and the other practices high-low pricing. There are also some differences in the type of data from the two retailers.

The first data set (Retailer 1) consists of point-of-sale market-basket data from 20 stores that belong to a major grocery retail chain. The data include sales, price, and profit information on all items in every market basket purchased during a four-week period in 1998. The data result in records for items from 291 product categories bought in 801,821 market baskets. Of the product categories, 123 are in the produce, meats, deli, seafood, and bakery departments, in which most products are unbranded and not categorized into store and national brands. Of the remaining 168 product categories, which are in the grocery, general merchandise, dairy, specialty, frozen foods, and health/beauty/cosmetics (HBC) departments, 115 have a store brand. The average market share of the store brand is 22%, which is in line with the national average store-brand share of about 20% (Dunne and Narasimhan 1999). The profit data from Retailer 1 include both gross margin (after promotional allowances linked to a specific product are accounted for) and net contribution margin (after direct product costs incurred by the retailer, including receiving, stocking, packaging, cashiering, bagging, and so on, are subtracted) of every item in every market basket. However, the data are organized at the market-basket level, not at the panel level, so we cannot assign baskets to individual households. Because pricing and retail margins differ across the 20 stores, we conducted our analysis at the store level for Retailer 1.

To study the determinants of retail margins, we combined the sales, price, and profit variables we obtained from the market-basket data with category penetration, average purchase cycle, and promotion activity from the Marketing Fact Book (Information Resources Inc. 1998); category advertising data from Leading National Advertisers' class/ brand summary; store-brand quality data collected by Hoch and Banerji (1993) and Sethuraman and Cole (1997); and distribution data collected in our own survey of stores that belong to competing grocery, drug, and mass-merchandiser chains.

The second data set (Retailer 2) is based on purchase data from the drug retailer's customer panel for all the stores in a market during a six-month period in 2000. The data include price and gross margin for each item purchased by the panel members in every shopping trip and contain records for 41,335 customers who made a total of 347,214 shopping trips. The total number of categories is 118, and the retailer offers a store brand in 70 categories. The average store-brand market share is approximately 23%. As does Retailer 1, Retailer 2 also has a store brand in the majority of categories, and the store-brand share is consistent with levels reported in previous research. The stores in this data set are all in a single pricing zone with identical prices and margin, so we aggregated data across stores for our analysis.

The profit data from Retailer 2 include gross margin of every item in every market basket. We do not have net contribution margin in this case. However, the data are organized at the panel level, so we were able to evaluate the profitability of a customer, not just the profitability of a market basket. We also calculated category penetration, average purchase cycle, and percentage sold on promotion from the panel data. Table 1 lists all the variables we used in our study as well as their definitions and the sources from which we obtained them.

Model Development and Hypotheses

The determinants of retail margins can be grouped into three sets: variables that reflect ( 1) manufacturer market power, ( 2) retailer market power, and ( 3) consumer and category characteristics. An overarching theme that underlies our hypotheses, especially for the first two sets of variables, is the relationship of market power to margins and the inverse relationship between manufacturer and retail margins (Steiner 1993). In brief, as the manufacturer's consumer franchise and market power increase, wholesale prices and manufacturer margins increase. However, retailers cannot increase prices on the products with a strong consumer franchise because consumers may switch stores to buy their preferred brand at a better price. As a result, retail margins decrease. In contrast, if consumers are more likely to switch brands within a store than to switch stores, retail market power increases, wholesale prices are negotiated down, and manufacturer margins decrease. However, lower retail competition reduces the downward pressure on retail prices, and retail margins increase as a result.

We discuss the individual variables in our model and our expectations about their effect on retailers' gross and net contribution margins from national and store brands. Note that we expect that only the variables that influence direct product costs will have significantly different effects on gross margins than on net contribution margins. Table 2 summarizes our hypotheses about all the variables in the model, in the order that we discuss them subsequently.

Manufacturer market power. As the concentration of national-brand manufacturers increases, their market power also increases (Scherer and Ross 1990). Therefore, we expect retailers to take lower margins on national brands in categories in which the Herfindahl index, a standard measure of concentration, is high. Furthermore, because higher manufacturer power leads not only to higher wholesale prices but also to lower retail prices for national brands, retailers must reduce prices on store brands to preserve a sufficient price differential with national brands. Therefore, we expect retailers also to take lower margins on store brands as the Herfindahl index increases.

Beginning with the pioneering work of Steiner (1973, 1978), researchers have shown that national brands that manufacturers advertise heavily are sold at higher wholesale prices and lower retail margins than are less advertised brands (Albion and Farris 1981, 1987; Lal and Narasimhan 1996; Steiner 1993). This phenomenon is because advertising decreases consumers' propensity to switch to another brand within the store and increases their propensity to switch stores for their preferred brand, thus giving the manufacturer leverage over retailers. If the same phenomenon applies across product categories, we expect retailers to take lower margins on national brands in heavily advertised categories. To compete effectively in categories with high consumer pull for national brands, we expect retailers to take lower margins on store brands as well. We also expect both national brands and store brands to be sold at lower retail margins in heavily promoted categories because retailers must price more competitively in such categories (Gerstner, Hess, and Holthausen 1994; Steiner 1984).

Finally, because there is no reason to believe that any of the manufacturer power-related variables affect direct product costs, we expect them to have similar effects on gross and net contribution margins.

Retailer market power. The greater the number of retailers who distribute the category, the greater is the competition between retailers and the lower is their market power. As a result, there will be greater pressure on retailers to keep their national brand prices competitive (Shankar and Bolton 2003). Therefore, we expect retail margins on national brands to be lower for categories that are widely distributed at competing chains. As the number of competing retailers that distribute a store brand in the category increases, the competition among retailers should also increase and put downward pressure on retail prices and margins. However, there is also a counterargument that store brands engender store loyalty and therefore give retailers some monopoly power because competing retailers do not carry the same store brand. This may soften store-brand price competition among retailers and enable retailers to take higher margins on store brands. We expect that wider store-brand distribution will make retailers compete more aggressively on store-brand prices, leading to a negative relationship between store-brand distribution and retail margins on both store brands and national brands.

Because store brands provide a means of differentiation and thus market power for retailers, retailers should be able to negotiate lower prices from manufacturers and earn higher margins in categories with a strong store-brand threat (Mills 1995; Morton and Zettelmeyer 2000; Narasimhan and Wilcox 1998). As we noted previously, some researchers have argued that a store brand can be a credible threat even if its share is low. However, we hypothesize that national-brand retail margins are higher in categories in which store-brand share is high because we believe high share is definitely a threat whereas low share may not be a credible threat. We also expect that store-brand margins are higher in categories with high store-brand share because high store-brand share reflects the strength of store-brand demand compared with national-brand demand (e.g., Raju, Sethuraman, and Dhar 1995).

Basic microeconomic theory predicts an inverse relationship between optimal gross margin and price elasticity for price setters. The price elasticity of demand that retailers face varies across categories, partly as a result of varying intensity in retail competition across categories. Although we have already included distribution, which is a key variable reflecting the level of competition retailers face, we also directly included category price elasticity. We expect that retailers will take lower margins on both national brands and store brands in categories in which the price elasticity of the consumer demand they face is high. Finally, because there is no reason to believe that the retailer power variables influence direct product costs, we expect them to have similar effects on gross and net margins.

Consumer and category characteristics. Retailers may take different margins on categories, depending on how many consumers buy the category, what the typical purchase amount is, and how often it is purchased. Their direct product costs may also differ with these variables because purchase cycles and the number of transactions can influence the cost of warehousing, packaging, stocking, cashiering, and so on. There is no strong theoretical basis for a priori hypotheses about the effects of the variables on retail margins; however, some prior empirical research does suggest that retailers are willing to take lower percentage gross margins in categories in which the purchase amount or penetration is high (Lal and Narasimhan 1996; Narasimhan and Wilcox 1998). In contrast, retailers may take higher gross margins in infrequently purchased categories because there are fewer opportunities for consumers to compare and learn about prices and for retailers to make money. Thus, we expect penetration and purchase amount to be negatively associated with gross margins on both national brands and store brands, and we expect purchase cycle to be positively associated with gross margins on both national brands and store brands. Furthermore, because direct product costs such as storage, stocking, and cashiering are likely to be lower as a percentage of sales in categories with large transaction amounts, we expect purchase amount to be positively associated with net margins on national and store brands. It is difficult to predict how the different components of direct product costs net out for high penetration and long purchase cycle categories, so we do not hypothesize a specific relationship between these variables and net margins.

To control for other unobserved differences across categories, we included dummy variables for the major departments represented in our sample, though we do not have any a priori hypotheses about their coefficients. Specifically, for Retailer 1, we included dummy variables for perishable foods and HBC categories because most of the remaining categories are in the grocery department. For Retailer 2, we only needed an HBC dummy variable because most of the other categories are in the general merchandise department. Finally, to control for Retailer 1's store-level differences in pricing and consequent margins that may be due to factors such as local competition, demographics, and so on, we included 19 store-specific dummy variables. As we noted previously, Retailer 2's stores are all from a single market and pricing zone, with identical prices and margins, so we aggregated across stores and did not need store-specific dummy variables.

Store-brand share model. Retailers may expend more effort to obtain higher store-brand share in categories in which gross margins are higher, and they may raise selling prices of national brands to increase store-brand share. This suggests that there may be reverse causality in the relationship between store-brand share and retail margins from national brands. To avoid any inconsistency due to this endogeneity, we specified a model for store-brand share and estimated the resultant simultaneous system by using three-stage least squares (3SLS). We modeled store-brand market share as a function of the category characteristics that Hoch and Banerji (1993) find to be significant determinants of store-brand share: number of manufacturers, total manufacturer advertising, dollar category volume, retail margin, average quality, and distribution of store brands. We included retail margin from national brands because reverse causality may be driven primarily by margin on national brands, as we noted previously. In addition, we included both the price elasticity of the category, because we expect the store brand to do better in categories in which consumers are more sensitive to price, and the ratio of store-brand to national-brand price, because store-brand share will increase as its price advantage increases. We do not discuss the store-brand share model in detail because it is not central to our research objectives and is directly based on prior research. We simply used it to obtain consistent estimates of the coefficients in the profit margin equations.

The simultaneous model. The previous discussion leads to the following simultaneous system for retail margins on both store brand and national brands and on store-brand share:( n1)

( 1) Margin[subkij] = γ[subk0] + γ[subk1] Herfind [subij] + γ[subk2]Advtg[subi] + γ[subk3]Deal[subi]

+ γ[subk4]Depth[subi] + γ[subk5]Catdist[subi] + γ[subk6]Sbdist[subi]

+ γ[subk7]Catelast[subi] + γ[subk8]Penet[subi] + γ[subk9]Puramt[subij]

+ γ[subk10]Purcycle[subi] + gamma;[subk11]Perish[subi] + γ[subk12]HBC[subi]

+ γ[subk13]Sdum1[subj] + ... + γ[subk31]sdum19[subj]

+ β[subk1]Sbshare[sub1j] + ε[subkij], and

( 2) Sbshare[subij] = γ[sub50] + γ[sub51]Numman[subij] + γ[sub52]Advtg[subi] + γ[sub53]Catvold[subij]

+ γ[sub54]SBdist[subi] + γ[sub55]Qual[subi] + γ[sub56]Catelast[subi]

+ γ[sub57]SDum1[subij] + ... + γ[sub525]Sdum1[subij]

+ β[sub51]Margin[sub1ij] + β[sub52]Pricerat[subij] + ε[sub5ij],

where k takes values of 1 and 2 for the retailer's gross margins from national and store brands and values of 3 and 4 for corresponding net margins.( n2) All the variables are defined in Table 1.

The system is identified because the margin equation contains seven variables that do not appear in the store-brand share equation and the store-brand share equation contains three variables that do not appear in the margin equations. We focus on the three variables excluded from the margin equations. First, note that three components of category volume (i.e., penetration, purchase amount, and purchase cycle) are included in the margin equation. As we discussed previously, the three components may have different effects on margin, and thus we included them separately. In contrast, there is no reason to expect differential effects on store-brand share, so the composite category volume variable is included in that equation (Hoch and Banerji 1993). Second, the number of manufacturers is not included in the margin equation, but we included the Herfindahl index, which is a well-established measure of manufacturer concentration. Third, the quality variable is a nationwide, general measure of the perceived quality of the best store brands in the country, not the quality of the specific retailer's store brands (see Table 1). By using the store-brand share variable, which is a direct measure that is specific to the retailer, we captured the increased leverage that strong store brands provide over manufacturers. We do not expect general, nationwide perceptions of store brands to affect the margins of a particular retailer over and above their effect on that retailer's store-brand share.

Model Estimates

We estimated the model for each retailer by using categories for which we had information on all relevant variables. We have complete data on 75 of the 115 categories for Retailer 1 and 45 of the 84 categories for Retailer 2. Correlation matrices of the variables in the margin models for the two retailers are provided in Tables 3 and 4, and 3SLS estimates of the margin equations for both retailers are provided in Table 5.( n3)

Manufacturer market power. Almost all our hypotheses about the effects of the variables pertaining to manufacturer market power are supported. The relationship of margins to the Herfindahl index, deal frequency, and deal depth was as we expected. The coefficients of the Herfindahl index and deal depth are significantly negative in five of the six regressions, and the coefficient of deal frequency is significantly negative in all six regressions. Category advertising has the expected negative effect on both retailers' national-brand margins, but, contrary to expectations, Retailer 1's margin on store brands increases significantly with advertising. Perhaps national brands in heavily advertised categories not only have higher wholesale prices but also have somewhat higher retail prices, thus enabling the retailer to charge higher prices for the store brand and still maintain a price advantage over national brands.

Retailer market power. Our hypotheses about the effects of the variables pertaining to retailer market power are also well supported in both data sets. As we expected, store-brand distribution has a negative relationship to both retailers' national-brand and store-brand margins. Heavy distribution of store brands increases price competitiveness of store brands instead of providing monopoly power to retailers. In addition, store-brand market share has the expected positive coefficient in five of the six regressions. The magnitude of the standardized coefficient of store-brand share in the national-brand margin model suggests that store-brand share is one of the most important drivers of retail margins on national brands. For Retailer 1, price elasticity is negatively associated with both national-brand and store-brand margins, though its coefficient is not statistically significant for Retailer 2. Contrary to our hypothesis, category distribution is not significant in most of the regressions. This may be due to multicollinearity, because the correlation between category and store-brand distribution is .67.( n4)

Consumer and category characteristics. We had only a few a priori hypotheses about the effects of consumer and category characteristics. As we expected, purchase cycle has a positive effect on national-brand and store-brand margins for both retailers. Purchase amount also has the expected positive effect on net contribution margin and the expected negative effect on gross margin, at least for the store brand of Retailer 2. The coefficient of penetration does not support our hypothesis: It is positive for Retailer 1 and insignificant for Retailer 2.

Summary. Table 6 summarizes our results for both retailers. For each of our a priori hypotheses, it depicts whether or not the hypothesis was supported. The symbol "check" denotes that the estimated coefficient is significant and of the hypothesized sign, and the symbol "X" denotes that the estimated coefficient is significant but not of the hypothesized sign. As the large number of "checks" show, most of the variables pertaining to manufacturer, retailer, and consumer and category characteristics influence retail profit margins on national brands and store brands as we hypothesized. The pattern of coefficients is similar for both retailers, though statistical significance is weaker for Retailer 2 because of its small sample size. This similarity provides important convergent validity for our results. Particularly notable is our finding that after we controlled for other key determinants and endogeneity of store-brand share, high store-brand share enables retailers to obtain significantly higher percentage margins on national brands. We find this result for both data sets in our analysis, despite substantial differences in the type of retailers, the product categories they sell, and their pricing and promotion policy.

The Percentage Margin Differential

Having developed some important insights into the determinants of retailers' percentage margins on national brands and store brands, we now examine the differential between the two. Table 7 reports the median ratio of store-brand to national-brand margins for the two retailers across all categories and for the retailers' major departments.

The first row of Table 7 confirms that on average across all the categories, not only percentage gross margins but also percentage net margins are substantially higher for the store brand than for national brands. The numbers for both retailers are consistent with the 20%-25% gross margin advantage for store brands that is noted in the literature (Hoch and Banerji 1993). The numbers are higher than the 16% margin advantage that Albion and Farris (1987) find for unadvertised versus advertised brands, but that is not surprising because only approximately 50% of the unadvertised brands in their sample were store brands. The remaining rows of Table 7 show that the ratio varies substantially across departments. For example, the store brand--national brand margin ratio is high in the HBC department for both retailers, likely because HBC contains strong brand names with heavy advertising that can command substantially higher wholesale prices from retailers. In contrast, Retailer 1 does not have a store-brand margin advantage in its general merchandise department, which contains many commodity categories, such as pet supplies, shoe care, and stationery. The wholesale prices of national brands in this department are probably quite close to the competitive wholesale prices of store brands, thus bringing the retailers' margins on national and store brands closer together.

Cross-Category Determinants of the Percentage Margin Differential

To examine the variations in the percentage margin differential across categories more systematically, we estimated a 3SLS model with the same explanatory variables as in the previous section, but with the retailer's margin on the store brand minus the margin on national brands as the dependent variable. Note that any variables that influence the margin differential do so by influencing one or both of its components. The (unstandardized) effect of any variable on the differential will be equal to its effect on store-brand margin minus its effect on national-brand margin (Farris, Parry, and Ailawadi 1992). Thus, only variables that have a substantially different influence on the retailer's store-brand margin rather than its national-brand margin will have a significant impact on the margin differential.

Table 8 summarizes the 3SLS coefficient estimates of the margin differential model for both retailers. As can be predicted from Table 5, many of the variables that pertain to manufacturer and retailer power have a significant influence on the margin differential because their impact on store-brand margin is different from their impact on national-brand margin. We discuss the most important influences here. First, category advertising significantly increases the margin differential because retailers must take lower margins on national brands in heavily advertised categories, but their store-brand margins do not appear to suffer. Second, the depth of deals in the category also increases the differential, probably because the deep discounts on national brands hurt the retailer's national-brand margins, and store brands are not discounted heavily from their everyday low prices. Third, store-brand distribution significantly reduces the margin differential. It appears that retailers need to price their store brands even more competitively than national brands when several competing retailers also offer a store brand. They do not seem to benefit from any local monopoly on their particular store brand. Fourth, store-brand share also reduces the margin differential. The retailer is able to obtain lower wholesale prices and therefore earn higher margins on national brands as store-brand share increases. As a result, the store-brand margin advantage becomes smaller.

The Dollar Margin Differential

As we noted in the first section, that retailers earn higher percentage margins on store brands than on national brands does not guarantee that their dollar margins on store brands will also be higher. For the dollar margins on national brands and store brands to be equal, the ratio of store-brand percentage margin to national-brand percentage margin must be equal to the ratio of national-brand retail price to store-brand retail price. Given the substantially lower price at which store brands sell, the retailer's dollar margin may be lower on store brands if the percentage margin advantage is low.

Table 9 summarizes the ratio of store-brand to national-brand prices and dollar margins for both retailers.( n5) The contrast between Table 7 and Table 9 is noteworthy. Although both retailers have similar store-brand percentage margin advantages, Retailer 2 has a store-brand dollar margin advantage but Retailer 1 does not. On average, Retailer 1's gross dollar margin per unit from store brands is lower than that from national brands in all the departments except HBC. The same is true of net contribution dollar margin. As Table 9 shows, for Retailer 1, the percentage margin advantage that the store brand offers does not make up for the substantially lower retail price of the store brand. Thus, from a perspective of dollar margin per unit, store brands do not necessarily offer retailers an advantage unless they are able to keep a relatively low price differential between the store brand and national brands.

Our finding that a retailer's dollar margin per unit may be lower on store brands than on national brands suggests that, all else being equal, a customer who buys store brands contributes less to the total dollar profit of a retailer than does a customer who buys national brands. As we discussed previously, all else may not be equal. A customer's contribution to the total dollar profit of the retailer depends not only on the dollar margin per unit of purchased items but also on the total number and type of items bought. In this section, we explore how the profitability of heavy store-brand users compares with that of light users or nonusers.

Method for Profitability Comparison

To perform the analysis for the market-basket data for Retailer 1, we used 646,549 market baskets that had nonmissing profitability data and that included at least one item from the 115 categories in which a store brand was offered. For each market basket, we computed the percentage of items purchased from these 115 categories that are the store brand. This procedure accounts for the availability of store brands in the categories represented in the basket and ensures that a large basket is not characterized as having low store-brand share simply because it contains many items from categories that do not have a store brand. Approximately 40% of the baskets do not contain any store-brand items, and in slightly more than 25% of the market baskets, store-brand items constitute less than 35% of the total number of items in which a store brand could have been bought. We used cutoffs of 0% and 35% to split the baskets into zero, low, and high store-brand share baskets, and then we compared mean profitability across the three groups of baskets. As we noted previously, we did not have panel data for Retailer 1 and thus could not identify baskets from the same household. The data provide some insights at the basket level but do not speak to the total profitability of a customer.

However, for Retailer 2 we did have panel data, so we could directly examine the relationship between customer-level profit and store-brand purchasing levels. We calculated the share of each customer's purchases that were store-brand items, and using the same cutoffs as we did for Retailer 1, we compared profit across groups of customers with zero, low, and high share.

We conducted this exploratory but less constrained comparison of groups instead of estimating a continuous relationship between store-brand share and market-basket/ customer profitability because prior research does not enable us to predict a specific kind of relationship.

Results of Profitability Comparison

Table 10 provides the mean values of four profit measures across the three groups of market baskets for Retailer 1 and the three groups of customers for Retailer 2: total gross margin dollars, total contribution margin dollars, total gross margin as a percentage of sales, and total contribution margin as a percentage of sales. It also provides the mean size of market baskets and customer sales in terms of both dollar value and number of items. Given the large sample sizes in both cases, group differences are statistically significant for all the variables, so we focus on managerial importance in the insights we draw from this table.

Percentage margin differences. For Retailer 1, there is little managerially significant difference across zero, low, and high store-brand baskets in either gross or net contribution margins as a percentage of sales. Given that percentage margins are higher for store brands than for national brands, we expected that the average would be somewhat higher in market baskets with a high share of store brands. The small difference suggests that market baskets with a high share of store brands contain other products that have somewhat lower percentage gross margins than average. In contrast, for Retailer 2, the average percentage gross margin from heavy users of store brands is much higher than from nonusers. Thus, there is no evidence that heavy store-brand users disproportionately buy other low-percentage-margin products from this retailer to offset their higher-percentage-margin store-brand purchases.

Dollar margin differences. Average dollar profit varies substantially across the groups for both retailers. There is an inverted U-shaped relationship between store-brand share and profit. For Retailer 1, the dollar profit of baskets with low store-brand share is the highest; it is almost twice that of high store-brand baskets and more than three times that of zero store-brand baskets. This is because zero and high store-brand baskets are much smaller than low store-brand baskets. The inverted U-shaped relationship also holds when we examined groups of customers, not just groups of baskets, for Retailer 2. The dollar profit is highest for the light store-brand users, dropping by roughly half among the heavy store-brand users and by much more among nonusers. The difference in dollar profit across customer groups is due to heavy store-brand users making fewer trips and buying much less from the retailer than light store-brand users do; nonusers make the fewest trips and buy the least.

Why is the profitability of customers who do not buy store brands at all so small? It would appear that they are not loyal to the retailer. Not only the size of their average market basket but also their total number of trips and their total purchases from the retailer are small. This is consistent with research that shows a positive relationship between store-brand use and store loyalty (Ailawadi, Neslin, and Gedenk 2001; Corstjens and Lal 2000). Perhaps the unwillingness of the customers to buy the store brand at all is a reflection of their relatively weak perception of the retailer.

Why is the profitability of heavy store-brand users also small? They do not just make more frequent trips to the same store and buy less per trip. This is evidenced by not only the market baskets of Retailer 1 but also the total customer profit of Retailer 2 showing the same pattern. Heavy store-brand users may ( 1) operate under financial constraints and simply buy and consume less, ( 2) have smaller families and therefore fewer requirements, or ( 3) shop at multiple stores and therefore contribute less to the sales and profit of any single retailer. Prior research provides strong evidence that heavy store-brand users operate under financial constraints and are price conscious (Ailawadi, Neslin, and Gedenk 2001; Batra and Sinha 2000; Richardson, Jain, and Dick 1996), in support of the first explanation. The limited demographic data that are available on the customers of Retailer 2 also provide some support for the second explanation. The last five rows of Table 10 display the demographic profile of the three groups of Retailer 2's customers and indicate that, in general, nonusers and light users are much more similar to each other than they are to heavy users. This is consistent with our belief that the factors underlying fewer purchases by heavy store-brand users are quite different from the factors underlying fewer purchases by non-store-brand users. Nonusers of store brands are not significantly different from light users in terms of gender, household size, or age, so it is difficult to argue that they have fewer requirements. It is more likely that they are the least loyal to the retailer. Heavy users of store brands are older and have smaller households than do light users or nonusers, suggesting that they may have fewer requirements.

However, it is unlikely that differences in requirements completely explain the substantial difference in expenditure between light and heavy store-brand users. Perhaps heavy store-brand users shop and "cherry-pick" at multiple stores and therefore buy less from any one retailer. This is consistent with a Food Marketing Institute (1994) study that found heavy store-brand buyers at several different stores. Heavy store-brand buyers may buy store brands wherever they shop because they consider them a good value (i.e., they may be loyal to store brands in general, not to the store brand of a particular retailer). This appears to contradict the positive relationship between store-brand use and store loyalty that Ailawadi, Gedenk, and Neslin (2001) and Corstjens and Lal (2000) observe. However, the positive association these researchers find in their linear models may be driven by the higher loyalty of store-brand users than nonusers rather than heavy store-brand users versus light users. Our analysis suggests a nonlinear relationship between store-brand use and store loyalty.

In this article, we developed an econometric model of the percentage margins that retailers earn on national brands and store brands. We used data from two different retailers to estimate our model and to identify the key factors that explain why the margins vary across categories. This analysis provided several insights into the role of variables such as manufacturer concentration, advertising, and distribution in decreasing retail margins on national brands and store brands. The analysis also resolved a debate about the effect of store-brand share on retail margins from national brands and showed that the store-brand share effect is significantly positive. It also suggests some avenues for further research. First, although we accounted for the endogeneity of store-brand share by using 3SLS, we were limited by an essentially cross-sectional data set. Longitudinal data would allow for more flexibility in selecting instrumental variables and in examining how margins change with store-brand share. Similarly, a "before--after" quasi experiment that examines how national-brand margins change after a store brand is introduced would provide a convincing test of the direction of causality (e.g., Pauwels and Srinivasan 2003). Second, there may be interactions between some of the determinants of retail margins in our model. We tested for an interaction between advertising and distribution (Abela and Farris 1999) and between store-brand quality and store-brand share (Narasimhan and Wilcox 1998), and we found that they were not significant. However, there may be other interactions that further research should consider. Third, both self-and cross-price elasticity determine product-line pricing and margins, according to economic theory. Although we included the category price elasticity in our model, further research should also account for the self-and cross-price elasticities of national and store brands. Fourth, our interest was in distinguishing between retail margins on national brands and store brands, but margins on individual national brands also vary substantially and should be studied. It would be worthwhile to integrate such analysis with research on retail price elasticities and retailer pricing strategies (e.g., Shankar and Bolton 2003).

Next, we examined the impact of store-brand share on retail margins in more detail by answering three hitherto unaddressed questions: ( 1) Are percentage margins on store brands higher than on national brands, and why does this margin advantage vary across product categories? ( 2) Does the store-brand advantage hold when we consider dollar margin per unit? and ( 3) How much do heavy store-brand users, compared with light users and nonusers of store brands, contribute to the total dollar profit of a retailer? On average, we found that percentage margins on store brands are significantly higher than on national brands; thus, all else being equal, category percentage margins increase with store-brand share. However, there are substantial variations across categories; the store-brand margin advantage is greater in heavily advertised categories and smaller in categories with widely available store brands and high store-brand share. Furthermore, the store-brand margin advantage does not necessarily hold when we consider dollar margin, because though store brands earn higher percentage margins, they are sold at substantially lower retail prices than national brands are. Thus, retailers' rushing to increase store brands across categories simply because of their higher percentage margins may not be a good idea. Retailers should expand their store-brand offering carefully and recognize that the biggest margin rewards are in categories in which competition from other retailers is relatively small and national brands are heavily advertised. Retailers should also try to position the store brand on reasonable quality, not just on low price, because having to compete heavily on price erodes store brands' dollar margin advantage.

Our findings about the profitability of heavy store-brand users also send a cautionary message to retailers. There are two pieces of "good news" here. First, heavy users do not selectively buy other items that are significantly less profitable on a percentage basis for the retailer. Second, customers who buy at least some store-brand items contribute much more to the sales and profit of the retailer than do customers who do not buy any store-brand items. However, heavy store-brand buyers buy significantly fewer items in total and therefore contribute much less to the dollar profit of the retailer than do light store-brand buyers. This is a notable finding that deserves further examination because it sheds light on the profitability of customers attracted by store brands. Existing research and our preliminary analysis that used the panel data from Retailer 2 suggest that heavy store-brand buyers have fewer requirements either because they are financially constrained, buy less, and consume less or because they have smaller households. However, it may also be that they shop at multiple stores and are loyal to store brands in general, not to the store brand of a particular store. Further research that combines more detailed demographic and psychographic variables with panel-purchase and profit data from multiple retailers in a market is needed in order to conclusively determine which explanation is true and to study the implications of each explanation for total retailer profit. It would also be useful to conduct a cross-retailer analysis that compares the profitability of retailers with relatively weak store brands with that of strong store brands.

Some authors have noted that increasing store-brand share beyond a certain point may not be wise. They argue that there may be a downside to high store-brand share because national brand manufacturers may no longer find it worthwhile to offer better prices and trade deals (e.g., Narasimhan and Wilcox 1998; Wellman 1997). For both retailers in our analysis, there was wide variation in store-brand share across categories, but we did not find a negative relationship between store-brand share and retail margins on national brands, even at high levels of store-brand share.

However, in the final analysis, it is not just percentage margin but also dollar profit that matters; here, there seems to be a downside to high store-brand share. First, at least in some cases, the dollar margin per unit from store brands is actually smaller than that from national brands, because store brands command much lower retail prices. Second, customers with medium rather than high levels of store-brand use appear to be the most profitable from a standpoint of total dollar profit. Retailers may find it worthwhile to offer store brands despite lower dollar margins, partly as leverage over manufacturers but partly to provide a lower price alternative to price-conscious consumers whose patronage they may otherwise lose. However, our analysis suggests that pushing store brands too much at the expense of national-brand offerings is not advisable. It is important that retailers retain a balance between store brands and national brands to attract and retain profitable customers who buy some store-brand items but not too many. This supports the view that national brands continue to be traffic builders and that reducing national brand choices may make a store less attractive to profitable shoppers (e.g., Farris and Ailawadi 1992; Johnson 1994). This is also consistent with Corstjens and Lal's (2000) finding that for a quality store-brand strategy to be profitable, there should be enough customers who buy national brands.

In closing, we recognize three limitations of our data. First, we study only two retailers, and though their store-brand shares are in line with nationwide averages, it would be valuable to test our model with data from more retailers. Still, we want to reiterate that the two retailers are different from each other in significant ways, and yet most of our empirical results from both data sets are similar. We believe this establishes the convergent validity of our results rather well. Second, manufacturers sometimes provide lump-sum trade allowances to retailers that are not linked to a particular item or even to a particular product category, and these data were not available to us. However, the lump-sum payments are made across multiple categories, so it is unlikely that they would systematically bias the coefficients in our category-level margin model. Third, the store-brand quality data we use are older than the data from the two retailers. Sethuraman and Cole (1997) conducted a survey that asked consumers to rate the quality of store brands compared with that of national brands, but we could not use their data in our analysis because they were gathered only for a subset of the categories we study. However, the correlation with the Hoch and Banerji (1993) data is .70 across the categories for which both sets of data are available, suggesting that the cross-category quality differences that Hoch and Banerji (1993) report are valid for our analysis.

In conclusion, we note that the interactions among manufacturers and retailers, retailer power and profit margins, and the role that store brands play are important issues in both the business and the academic press. We hope that this article has provided convincing answers to some questions about the impact of growing store-brand share on retailer margins and has provided food for thought and specific directions for further examination of other questions.

The authors thank two anonymous retail chains, Stephen Hoch, and Raj Sethuraman for providing the data used in the article. The authors also thank Peter Vishton and Pen-Che Ho for research computing assistance; Jeannie Newton and Adam Barrer for help with data collection; and Dick Bower, Paul Farris, Praveen Kopalle, Rajiv Lal, Don Lehmann, Scott Neslin, Koen Pauwels, Al Silk, and Bob Steiner for their helpful suggestions. The authors thank participants of the Tuck Marketing Seminar Series, the 2000 INFORMS Marketing Science Conference, the Marketing Science Institute, and the University of Texas at Dallas Marketing Seminar Series for their comments. Finally, the authors are grateful for the many valuable comments and suggestions from the four anonymous JM reviewers. This research was supported by the Tuck Associates Program.

(n1) The ratio of store-brand to national-brand price (Pricerat) is also treated as endogenous in this system.

(n2) For Retailer 2, k only takes values of 1 and 2 for gross margins from national and store brands. Furthermore, we aggregated data across stores, so there is no j subscript and there are no store-specific dummy variables in the model. Finally, because the deal and depth variables are determined by Retailer 2's own decisions and are not based on nationwide retail data, as in the case of Retailer 1, we treat the two variables as endogenous for Retailer 2.

(n3) Estimates of the store-brand share equation are available from the first author on request.

(n4) In addition, category distribution may not accurately reflect retail competition in some categories, such as light bulbs, in which category distribution may be high, but retailers carry different national brands to reduce competition.

(n5) It is a challenge to compute dollar margin per unit because individual items are measured in different sizes and units, not only across categories but also within a category. Furthermore, as is typical with scanner data, the size variable is not available in a numeric form that is easy to use in computations. We did our best to compute dollar margins per unit correctly. However, to minimize the impact of outliers that may have been a result of incomparable units of measurement, we sorted items in each category according to their dollar margin per unit and deleted the top and bottom 5% from this analysis.

Legend for Chart:

A - Variable
B - Definition
C - Source for Retailer 1
D - Source for Retailer 2

     A

            B

                C

                    D

Percentage gross margin

            Net selling price - cost from vendor /
            Net selling price

                Retailer POS data

                    Retailer panel data

Percentage net contribution margin

            Net selling price - cost from vendor - direct
            product cost / Net selling price

                Retailer POS data

                    Retailer panel data

Percentage margin for national
brands

            (Multiple lines cannot be converted
            in ASCII text)

                Retailer POS data

                    Retailer panel data

Dollar gross margin

            Net selling price - cost from vendor

                Retailer POS data

                    Retailer panel data

Dollar net contribution margin

            Net selling price - cost from
            vendor - direct product cost

                Retailer POS data

                    Retailer panel data

Dollar margin for national brands

            Median value of dollar margin for
            national brands

                Retailer POS data

                    Retailer panel data

Herfindahl index (Herfind)

            Sum of squared market shares of all
            manufacturers in the category

                Retailer POS data

                    Retailer panel data

Advertising (Advtg)

            Media advertising for product
            category (millions of dollars)

                Leading National Advertisers' class/brand
                summary

                    Leading National Advertisers' class/brand
                    summary

Deal frequency (Deal)

            Percentage of all sales of the category made
            on some type of deal

                Information
                Resources Inc.
                (1998)

                    Retailer panel data

Deal depth (Depth)

            Average percentage discount when the category
            is on some type of deal

                Information
                Resources Inc.
                (1998)

                    Retailer panel data

Category distribution (Catdist)

            Percentage of major grocery, drug, and
            mass-merchandise chains in the relevant
            geographical area that carry the product
            category

                Primary data (2002)

                    Primary data (2002)

Store-brand distribution (Sbdist)

            Percentage of major grocery, drug, and
            mass-merchandise chains in the relevant
            geographical area that carry a store brand
            in the category

                Primary data (2002)

                    Primary data (2002)

Store-brand share (Sbshare)

            Dollar sales of store brand /
            Total dollar sales of product category

                Retailer POS data

                    Retailer panel data

Category elasticity (Catelast)

            Average percentage increase in category sales
            due to a 15% price cut

                Narasimhan, Neslin, and Sen (1996); P-O-P
                Times (1991)

                    Narasimhan, Neslin, and Sen (1996); P-O-P
                    Times (1991)

Penetration (Penet)

            Percentage of households that buy the category

                Information
                Resources Inc.
                (1998)

                    Retailer panel data

Average purchase amount (Puramt)

            Average dollar amount of purchase when
            category is purchased (i.e., price per
            unit x number of units purchased)

                Retailer POS data

                    Retailer panel data

Average purchase cycle (Purcycle)

            Average number of days between consecutive
            purchases of the category

                Information
                Resources Inc.
                (1998)

                    Retailer panel data

Perishable product dummy (Perish)

            Equal to 1 if the category is perishable
            (i.e., frozen, refrigerated, or dairy foods);
            equal to 0 otherwise

                Retailer POS data

                    Retailer panel data

HBC dummy (HBC)

            Equal to 1 if the category is in the HBC
            department; equal to 0 otherwise

                Retailer POS data

                    Retailer panel data

Number of manufacturers (Numman)

            Number of manufacturers whose products are
            sold in the category

                Retailer POS data

                    Retailer panel data

Category dollar volume (Catvold)

            Category sales in dollars

                Retailer POS data

                    Retailer panel data

Average store-brand quality: 1 (Qual)

            Average response by retail experts to "How does
            the quality of the best private label supplier
            compare to the leading national brands in the
            product category?" (five-point scale)

                Hoch and Banerji (1993)

                    Hoch and Banerji (1993)

Average store-brand quality: 2

            Average response by consumers to "Please indicate
            your opinion about the quality of private labels
            when compared with the quality of national brands
            for each product category" (seven-point scale)

                Sethuraman and Cole (1997)

                    Sethuraman and Cole (1997)

Ratio of store-brand to national-brand
price (Pricerat)

            Retail selling price of store brand /
            Average retail selling price of
            national brands in the category

                Retailer POS data

                    Retailer panel data

Store dummies (Sdum 1-19)

            Equal to 1 if the observation is for that store;
            equal to 0 otherwise

                Retailer POS data

                    --

Notes: POS = point of sale.

Legend for Chart:

A - Independent Variables
B - Hypothesized Sign of Coefficient for Retail Margin National
    Brands Percentage Gross Margin
C - Hypothesized Sign of Coefficient for Retail Margin National
    Brands Percentage Contribution
D - Hypothesized Sign of Coefficient for Retail Margin Store
    Brands Percentage Gross Margin
E - Hypothesized Sign of Coefficient for Retail Margin Store
    Brands Percentage Contribution
F - Selected Literature Support for Hypotheses

    A                                B    C    D    E

                                     F

Manufacturer Market Power

Herfindahl                           -    -    -    -
index

                                     Scherer and Ross (1990)

Advertising                          -    -    -    -

                                     Albion and Farris (1981),
                                     Lal and Narasimhan (1996),
                                     Steiner (1973, 1978, 1993)

Percentage                           -    -    -    -
sold on deal

                                     Chevalier and Curhan (1976),
                                     Steiner (1984)

Deal depth                           -    -    -    -

                                     Chevalier and Curhan (1976),
                                     Gerstner, Hess, and
                                     Holthausen (1994), Steiner
                                     (1984)

Retailer Market Power

Category                             -    -    -    -
distribution

                                     Abela and Farris (1999),
                                     Scherer and Ross (1990)

Store-brand                          -    -    -    -
distribution

                                     Abela and Farris (1999),
                                     Scherer and Ross (1990)

Store-brand                          +    +    +    +
share

                                     Mills (1995)

Category price                       -    -    -    -
elasticity

                                     Albion and Farris (1987),
                                     Mansfield (1997)

Consumer/Category Characteristics

Category                             -    ?    -    ?
penetration

                                     Chiang and Wilcox (1997),
                                     Lal and Narasimhan (1996)

Purchase cycle                       +    ?    +    ?

                                     Narasimhan and Wilcox (1998)

Purchase                             -    +    -    +
amount

                                     Lal and Narasimhan (1996),
                                     Narasimhan and
                                     Wilcox (1998)

Legend for Chart:

B - Nbgm
C - Nbcm
D - Sbgm
E - Sbcm
F - Herfind
G - Advtg
H - Deal
I - Depth
J - Catdist
K - Sbdist
L - Sbshare
M - Catelast
N - Penet
O - Purcyc
P - Puramt
Q - Perish

 A            B      C      D      E      F      G      H      I
              J      K      L      M      N      O      P      Q

Nbgm           1

Nbcm         .91      1

Sbgm         .49    .49      1

Sbcm         .45    .51    .96      1

Herfind      .13    .07   -.16   -.15      1

Advtg       -.27   -.19    .09    .16   -.15      1

Deal        -.35   -.28   -.45   -.37    .04    .06      1

Depth       -.21   -.26   -.03   -.07   -.14   -.04    .14      1

Catdist     -.21   -.19    .03    .01   -.12    .21   -.18   -.08
               1

Sbdist      -.25   -.17   -.03    .01   -.21    .32   -.07   -.13
             .67      1

Sbshare      .26    .23   -.01   -.02    .39   -.22   -.06   -.20
            -.08    .01      1

Catelast    -.20   -.07   -.16   -.07   -.15    .17    .29   -.33
             .19    .31   -.05      1

Penet       -.08   -.09   -.28   -.26    .12   -.08    .39    .11
            -.00   -.09    .21    .13      1

Purcycle     .16    .25    .37    .40   -.15   -.07   -.30    .07
             .03    .11   -.13   -.04   -.56      1

Puramt      -.04    .18    .25    .38   -.02    .30   -.12   -.29
             .21    .29   -.16    .30   -.34    .22      1

Perish       .15    .16   -.08   -.07    .15   -.13    .22    .01
            -.81   -.50    .15   -.22    .07   -.28   -.14      1

HBC         -.10   -.03    .42    .42   -.28    .32   -.27    .21
             .32    .39   -.22   -.09   -.44    .41    .36   -.23

Legend for Chart:

B - Nbgm
C - Sbgm
D - Herfind
E - Advtg
F - Deal
G - Depth
H - Catdist
I - Sbdist
J - Sbshare
K - Catelast
L - Penet
M - Purcyc
N - Puramt

   A          B       C       D       E       F       G       H
                      I       J       K       L        M      N

Nbgm           1

Sbgm         .57       1

Herfind      .27    -.32       1

Advtg       -.34     .12     .07       1

Deal        -.36    -.38     .11     .02       1

Depth       -.19     .04    -.18     .06     .19       1

Catdist     -.11    -.02    -.09     .31    -.21     .06      1

Sbdist      -.28    -.05    -.16     .24     .01    -.06    .54
                       1

Sbshare      .36     .03     .22    -.17    -.03    -.24    .03
                    -.03       1

Catelast    -.16    -.21    -.21     .25     .34    -.25    .24
                     .34    -.03       1

Penet       -.11    -.09     .17    -.11     .27     .04    .02
                    -.14     .23     .17       1

Purcycle     .26     .28    -.11    -.02    -.22     .09    .05
                     .15    -.09     .02    -.59       1

Puramt       .02     .33     .06     .27    -.17    -.34    .27
                     .26    -.07     .36    -.41     .26      1

HBC          .04     .09    -.26     .38    -.36     .16    .29
                     .44    -.19    -.05    -.37     .38    .32

Legend for Chart:

A - Independent Variable
B - Standardized Coefficient in Equation for Retailer 1 National
    Brands Percentage Gross Margin
C - Standardized Coefficient in Equation for Retailer 1 National
    Brands Percentage Contribution
D - Standardized Coefficient in Equation for Retailer 1 Store
    Brand Percentage Gross Margin
E - Standardized Coefficient in Equation for Retailer 1 Store
    Brand Percentage Contribution
F - Standardized Coefficient in Equation for Retailer 2 National
    Brands Percentage Gross Margin
G - Standardized Coefficient in Equation for Retailer 2 Store
    Brand Percentage Gross Margin

     A                              B                C

                                    D                E

                                    F                G

Herfindahl index                 -.004            -.059(***)
                                (-.15)          (-2.16)

                                 -.159(***)       -.143(***)
                               (-5.89)          (-5.42)

                                 -.284(***)       -.417(***)
                               (-2.27)          (-2.77)

Advertising                      -.059(***)       -.044(***)
                               (-2.28)          (-1.76)

                                  .103(***)        .131(***)
                                (4.28)           (5.57)

                                 -.315(***)        .082
                               (-2.83)            (.11)

Percentage sold on deal          -.252(***)       -.192(***)
                               (-9.57)          (-7.48)

                                 -.286(***)       -.231(***)
                              (-11.15)          (-9.18)

                                 -.523(***)       -.293(***)
                                (3.62)          (-2.81)

Deal depth                       -.191(***)       -.212(***)
                               (-7.23)          (-8.15)

                                 -.088(***)       -.116(***)
                               (-3.30)          (-4.44)

                                 -.172(**)        -.007
                               (-1.82)           (-.08)

Category distribution            -.094(***)       -.010
                               (-2.06)           (-.22)

                                 -.026            -.121(***)
                                (-.57)          (-2.68)

                                 -.087            -.133
                               (-1.12)           (1.09)

Store-brand distribution         -.179(***)       -.166(***)
                               (-5.24)          (-5.03)

                                 -.305(***)       -.240(***)
                               (-9.44)          (-7.57)

                                 -.308(***)       -.434(***)
                               (-3.62)          (-3.97)

Store-brand share                 .424(***)        .353(***)
                                (8.34)           (7.21)

                                  .233(***)        .121(***)
                                (4.91)           (2.61)

                                  .312(***)        .194
                                (2.74)           (1.06)

Price elasticity                 -.064(***)       -.052(**)
                               (-2.34)          (-1.96)

                                 -.067(***)       -.079(***)
                               (-2.53)          (-3.08)

                                 -.118             .172
                                (-.77)            (.73)

Category penetration              .105(***)        .250(***)
                                (3.60)           (8.73)

                                  .144(***)        .243(***)
                                (4.87)           (8.34)

                                 -.044            -.026
                                (-.17)           (-.31)

Purchase cycle                    .201(***)        .377(***)
                                (6.98)          (13.36)

                                  .211(***)        .295(***)
                                (7.30)          (10.36)

                                  .186(*)          .341(**)
                                (1.61)           (1.88)

Purchase amount                  -.034             .207(***)
                               (-1.34)           (8.21)

                                  .146(***)        .298(***)
                                (5.57)          (11.52)

                                 -.042            -.231(**)
                                (-.64)          (-1.89)

Perishable foods                  .048             .200(***)
                                (1.19)           (4.96)

                                 -.038            -.059
                                (-.90)          (-1.42)

                                   --               --

HBC                               .020            -.004
                                 (.70)           (-.15)

                                  .367(***)        .315(***)
                               (12.28)          (10.71)

                                  .033            -.062
                                 (.82)           (-.68)

(*) p < .15.
(**) p < .10.
(***) p < .05.

Notes: There are 19 store dummy variables in Retailer 1's model;
the coefficients are not reported here to conserve space. The
t-statistics are in parentheses.

Legend for Chart:

A - Independent Variable
B - Sign of Coefficient in Equation for National-Brand Margin
    Percentage Gross Margin Hypothesized
C - Sign of Coefficient in Equation for National-Brand Margin
    Percentage Gross Margin Retailer 1
D - Sign of Coefficient in Equation for National-Brand Margin
    Percentage Gross Margin Retailer 2
E - Sign of Coefficient in Equation for National-Brand Margin
    Percentage Contribution Hypothesized
F - Sign of Coefficient in Equation for National-Brand Margin
    Percentage Contribution Retailer 1
G - Sign of Coefficient in Equation for Store-Brand Margin
    Percentage Gross Margin Hypothesized
H - Sign of Coefficient in Equation for Store-Brand Margin
    Percentage Gross Margin Retailer 1
I - Sign of Coefficient in Equation for Store-Brand Margin
    Percentage Gross Margin Retailer 2
J - Sign of Coefficient in Equation for Store-Brand Margin
    Percentage Contribution Hypothesized
K - Sign of Coefficient in Equation for Store-Brand Margin
    Percentage Contribution Retailer 1

     A                               B     C       D     E   F
                                     G     H       I     J   K

Manufacturer Market Power

Herfindahl index                     -    N.S.    Y      -   Y
                                     -    Y       Y      -   Y

Advertising                          -    Y       Y      -   Y
                                     -    X       N.S.   -   X

Percentage sold on deal              -    Y       Y      -   Y
                                     -    Y       Y      -   Y

Deal depth                           -    Y       Y      -   Y
                                     -    Y       N.S.   -   Y

Retailer Market Power

Category distribution                -    Y       N.S.   -   N.S.
                                     -    N.S.    N.S.   -   Y

Store-brand distribution             -    Y       Y      -   Y
                                     -    Y       Y      -   Y

Store-brand share                    +    Y       Y      +   Y
                                     +    Y       N.S.   +   Y

Category elasticity                  -    Y       N.S.   -   Y
                                     -    Y       N.S.   -   Y

Consumer/Category Characteristics

Category penetration                 -    X       N.S.   ?   +
                                     -    X       N.S.   ?   +

Purchase cycle                       +    Y       Y      ?   +
                                     +    Y       Y      ?   +

Purchase amount                      -    N.S.    N.S.   +   Y
                                     -    X       Y      +   Y

Notes: Y = statistically significant and of hypothesized sign;
X = statistically significant but not of hypothesized sign;
N.S. = not significant.

Legend for Chart:

A - Group
B - Retailer 1: Median Ratio of Store-Brand Margin to National
    Brand Margin Number of Observations
C - Retailer 1: Median Ratio of Store-Brand Margin to National
    Brand Margin Percentage Gross Margin
D - Retailer 1: Median Ratio of Store-Brand Margin to National
    Brand Margin Percentage Net Margin
E - Retailer 2: Median Ratio of Store-Brand Margin to
    National-Brand Margin Number of Observations
F - Retailer 2: Median Ratio of Store-Brand Margin to
    National-Brand Margin Percentage Gross Margin

         A               B       C       D      E      F

Full sample             2063    1.27    1.27    84    1.29
Dairy products           184    1.10    1.06     0      --
Frozen foods             254    1.18    1.16     0      --
Specialty foods           89    1.05     .99     0      --
Grocery                 1026    1.30    1.30    11    1.40
General merchandise      161     .92     .74    37    1.26
HBC                      349    1.73    1.75    36    1.37

Legend for Chart:

A - Independent Variable
B - Retailer 1 Percentage Gross Margin Differential
C - Retailer 1 Percentage Net Contribution Differential
D - Retailer 2 Percentage Gross Margin Differential

      A                            B                C

                                                    D

Herfindahl index                 -.152(***)       -.087(***)
                               (-5.48)          (-2.99)

                                                  -.141(***)
                                                (-2.52)

Advertising                       .164(***)        .185(***)
                                (6.56)           (7.04)

                                                   .334(***)
                                                 (2.74)

Percentage sold on deal          -.080(***)       -.076(***)
                               (-3.06)          (-2.77)

                                                   .238(***)
                                                 (3.12)

Deal depth                        .074(***)        .062(***)
                                (2.74)           (2.22)

                                                   .158(**)
                                                 (1.77)

Category distribution             .060            -.118(***)
                                (1.29)          (-2.42)

                                                  -.048
                                                (-1.08)

Store-brand distribution         -.165(***)       -.108(***)
                               (-4.92)          (-3.07)

                                                  -.129(**)
                                                (-1.72)

Store-brand share                -.126(***)       -.186(***)
                               (-2.57)          (-3.60)

                                                  -.118(***)
                                                (-2.15)

Price elasticity                 -.016            -.042(*)
                                (-.59)          (-1.47)

                                                   .274(**)
                                                 (1.92)

Category penetration              .057(**)         .036
                                (1.91)           (1.14)

                                                   .021
                                                  (.17)

Purchase cycle                    .044(*)         -.022
                                (1.49)           (-.73)

                                                   .148(**)
                                                 (1.75)

Purchase amount                   .187(***)        .141(***)
                                (7.08)           (5.11)

                                                  -.206(*)
                                                (-1.53)

Perishable foods                 -.079(**)        -.239(***)
                               (-1.88)          (-5.43)

                                                    --

HBC                               .369(***)        .341(***)
                               (12.23)          (10.87)

                                                  -.089
                                                 (-.73)

(*) p < .15.
(**) p < .10.
(***) p < .05.

Notes: There are also 19 store dummy variables in Retailer 1's
models; t-statistics are in parentheses.

Legend for Chart:

A - Group
B - Median Ratio of Store Brand to National Brands Price per Unit
C - Median Ratio of Store Brand to National Brands Dollar Gross
    Margin per Unit
D - Median Ratio of Store Brand to National Brands Dollar Net
    Margin per Unit

    A                      B       C       D

Retailer 1

Full sample               .70     .87     .89
Dairy products            .77     .87     .89
Frozen foods              .69     .78     .77
Specialty foods           .62     .68     .69
Grocery                   .70     .90     .91
General merchandise       .59     .60     .57
HBC                       .62    1.18    1.26

Retailer 2

Full sample               .91    1.18      --
Grocery                   .82    1.15      --
General merchandise       .87    1.10      --
HBC                       .94    1.29      --

Legend for Chart:

A - Variable
B - Mean Value for Market Baskets Whose Store-Brand Share Is 0%
C - Mean Value for Market Baskets Whose Store-Brand Share Is <35%
D - Mean Value for Market Baskets Whose Store-Brand Share Is
    ≥35%

    A                                 B          C          D

Retailer 1

Percentage gross margin              27.6%      26.2%      27.4%
Contribution margin/sales            21.1%      19.9%      20.2%
Total gross margin dollars           $3.12      $9.95      $5.48
Total contribution dollars           $2.48      $7.67      $4.14
Total number of Items                 5.98      23.28      13.45
Total dollar amount                 $11.30     $38.54     $20.78
Percentage of market baskets         40.7%      27.6%      31.7%

Retailer 2

Percentage gross margin              27.4%      32.4%      36.8%
Total gross margin dollars          $15.88     $77.79     $48.35
Total number of items                16.27      80.17      46.26
Total dollar amount                 $54.68    $244.23    $128.53
Number of trips                       3.15      10.12       7.05
Percentage of customers              11.7%      77.1%      11.2%
Percentage male                      22.2%      22.5%      33.3%
Household size                         2.8        2.8        2.6
Age                                   47.4       46.6       53.1
Percentage bought on coupons          8.3%       8.7%       6.6%
Percentage bought on promotion       32.0%      32.1%      33.4%

Abela, Andrew V. and Paul W. Farris (1999), "Advertising and Competition," working paper, Darden School, University of Virginia.

Ailawadi, Kusum L., Donald R. Lehmann, and Scott A. Neslin (2001), "Market Response to a Major Policy Change in Marketing Mix: Learning from Procter & Gamble's Value Pricing Strategy," Journal of Marketing, 65 (January), 44-61.

------, Scott A. Neslin, and Karen Gedenk (2001), "Pursuing the Value-Conscious Consumer: Store Brands Versus National Brand Promotions," Journal of Marketing, 65 (January), 71-89.

Albion, Mark S. and Paul W. Farris (1981), The Advertising Controversy: Evidence on the Economic Effects of Advertising. Boston: Auburn House Publishing.

----- and ----- (1987), "Manufacturer Advertising and Retail Gross Margins," in Advances in Marketing and Public Policy, Vol. 1, Paul N. Bloom, ed. Greenwich, CT: JAI Press, 107-136.

Alpert, Frank H., Michael A. Kamins, and John L. Graham (1992), "An Examination of Reseller Buyer Attitudes Toward Order of Brand Entry," Journal of Marketing, 56 (July), 25-37.

Batra, Rajeev and Indrajit Sinha (2000), "Consumer-Level Factors Moderating the Success of Private Label Brands," Journal of Retailing, 76 (2), 175-91.

Besanko, David, Sachin Gupta, and Dipak Jain (1998), "Logit Demand Estimation Under Competitive Pricing Behavior: An Equilibrium Framework," Management Science, 44 (11), 1533-47.

Borin, Norm and Paul W. Farris (1990), "An Empirical Comparison of Direct Product Profit and Existing Measures of SKU Productivity," Journal of Retailing, 66 (3), 297-314.

Buzzell, Robert D., John A. Quelch, and Walter J. Salmon (1990), "The Costly Bargain of Trade Promotion," Harvard Business Review, 68 (March--April), 141--49.

Cannondale Associates (1996), Trade Promotion Spending and Merchandising: The New Paradigm. Evanston, IL: Cannondale Associates.

Chevalier, M. and Ronald Curhan (1976), "Retail Promotions as a Function of Trade Promotions: A Descriptive Analysis," Sloan Management Review, 18 (Fall), 19--32.

Chiang, Jeongwen and Ronald T. Wilcox (1997), "A Cross-Category Analysis of Shelf-Space Allocation, Product Variety, and Retail Margins," Marketing Letters, 8 (2), 183-91.

Chintagunta, Pradeep, Andre Bonfrer, and Inseong Song (2002), "Investigating the Effects of Store Brand Introduction on Retailer Demand and Pricing Behavior," Management Science, 48 (October), 1242-67.

Corstjens, Marcel and Rajiv Lal (2000), "Building Store Loyalty Through Store Brands," Journal of Marketing Research, 37 (Fall), 281-91.

Dhar, Sanjay and Stephen Hoch (1997), "Why Store Brand Penetration Varies by Retailer," Marketing Science, 16 (3), 208-227.

Discount Merchandiser (1996), "Revealing Private Thoughts," 36 (November), 58-62.

Dunne, David and Chakravarthi Narasimhan (1999), "The New Appeal of Private Labels," Harvard Business Review, 77 (May-June), 41-52.

Farris, Paul W. and Kusum L. Ailawadi (1992), "Retail Power: Monster or Mouse?" Journal of Retailing, 68 (4), 351-69.

-----, Mark Parry, and Kusum L. Ailawadi (1992), "Structural Analysis of Models with Composite Dependent Variables," Marketing Science, 11 (Winter), 73-94.

Food Marketing Institute (1994), Consumer Perspectives on National and Store Brands. Washington, DC: Research Department, Food Marketing Institute, 8-9.

Gerstner, Eitan, James D. Hess, and Duncan M. Holthausen (1994), "Price Discrimination Through a Distribution Channel: Theory and Evidence," American Economic Review, 84 (December), 1437-45.

Handy, Charles R. (1985), "Surveying Private and National Brand Prices," National Food Review, 29, 14-17.

Harrison, Dan (1999a), "PLMA Hears Why Private Label Aids Chains," Frozen Food Age, 47 (12), 44-45.

----- (1999b), "Strong Store Brands Add Edge to Buyer-Seller Tie," Frozen Food Age, 47 (12), 46-47.

Hoch, Stephen J. and Shumeet Banerji (1993), "When Do Private Labels Succeed?" Sloan Management Review, 34 (4), 57-67.

Information Resources Inc. (1998), Marketing Fact Book Annual Report, January-December. Chicago: Information Resources Inc.

Johnson, Jay L. (1994), "Private Labels: How High Is Up?" Discount Merchandiser, 34 (October), 30-37.

Kadiyali, Vrinda, Pradeep Chintagunta, and Naufel Vilcassim (2000), "A Manufacturer-Retailer Channel Interactions and Implications for Channel Power: An Empirical Investigation of Pricing in a Local Market," Marketing Science, 19 (2), 127-48.

Lal, Rajiv and Chakravarthi Narasimhan (1996), "The Inverse Relationship Between Manufacturer and Retailer Margins: A Theory," Marketing Science, 15 (2), 132-51.

Mansfield, Edwin (1997), Applied Microeconomics, 2d ed. New York: W.W. Norton.

Marketing News (1995), "Price, Quality Important for Private Labels Goods," (January 2), 24. Matthews, Ryan (1995), "Store Brand Competition Heats Up," Progressive Grocer, 74 (Special Issue), 19-20.

Messinger, Paul R. and Chakravarthi Narasimhan (1995), "Has Power Shifted in the Grocery Channel?" Marketing Science, 14 (2), 189-223.

Mills, David E. (1995), "Why Retailers Sell Private Labels," Journal of Economics and Management Strategy, 4 (Fall), 509-528.

Morton, Fiona Scott and Florian Zettelmeyer (2000), "The Strategic Positioning of Store Brands in Retailer-Manufacturer Bargaining," Working Paper No. 7712, National Bureau of Economic Research.

Narasimhan, Chakravarthi, Scott A. Neslin, and Subrata Sen (1996), "Promotional Elasticities and Category Characteristics," Journal of Marketing, 60 (April), 17-30.

----- and Ronald Wilcox (1998), "Private Labels and the Channel Relationship: A Cross-Category Analysis," Journal of Business, 71 (4), 573-600.

Pauwels, Koen and Shuba Srinivasan (2003), "Who Benefits from Store Brand Entry?" Marketing Science, 22.

P-O-P Times (1991), "Latest IRI Data Confirm the Effectiveness of P-O-P," (July-August), 37-42.

Quelch, John A. and David Harding (1996), "Brands Versus Private Labels: Fighting to Win," Harvard Business Review, 74 (January-February), 99-110.

Raju, Jagmohan S., Raj Sethuraman, and Sanjay Dhar (1995), "The Introduction and Performance of Store Brands," Management Science, 41 (June), 957-78.

Richardson, Paul, Arun K. Jain, and Alan S. Dick (1996), "Household Store Brand Proneness: A Framework," Journal of Retailing, 72 (2), 159-85.

Scherer, F.M. and David Ross (1990), Industrial Market Structure and Economic Performance, 3d ed. Boston: Houghton Mifflin.

Sethuraman, Raj and Catherine Cole (1997), "Why Do Consumers Pay More for National Brands Than for Store Brands?" Working Paper No. 97-126, Marketing Science Institute.

Shankar, Venkatesh and Ruth N. Bolton (2003), "An Empirical Analysis of Determinants of Retailer Pricing," Marketing Science, 22.

Steenkamp, Jan-Benedict E.M. and Marnik G. Dekimpe (1997), "The Increasing Power of Store Brands: Building Loyalty and Market Share," Long Range Planning, 30 (6), 917-30.

Steiner, Robert L. (1973), "Does Advertising Lower Consumer Prices?" Journal of Marketing, 37 (October), 19-26.

----- (1978), "A Dual Stage Approach to the Effects of Brand Advertising on Competition and Price," in Marketing and the Public Interest, John F. Cady, ed. Cambridge, MA: Marketing Science Institute.

----- (1984), "Basic Relationships in Consumer Goods Industries," Research in Marketing, 7, 165-207.

----- (1993), "The Inverse Association Between the Margins of Manufacturers and Retailers," Review of Industrial Organization, 8, 717-40.

Wellman, David (1997), "Souping Up Private Label," Supermarket Business, 52 (October), 13-20.

~~~~~~~~

By Kusum L. Ailawadi and Bari Harlam

Kusum L. Ailawadi is Associate Professor of Business Administration, Tuck School of Business, Dartmouth College (e-mail: kusum.l.ailawadi@dartmouth.edu).

Bari Harlam is Vice President, Marketing Intelligence, CVS Inc. (e-mail: baharlam@cvs.com).

An Exploratory Study of the Introduction of Online Reverse Auctions 

Buyers are increasingly turning to online reverse auctions in their negotiations with suppliers. How do these price competition mechanisms affect buyer-supplier relationships? The author considers this question in the context of a quasi experiment involving six online reverse auctions conducted in the supply base of a major industrial buyer. The results indicate that these auctions increase both new and current suppliers' beliefs that buyers act opportunistically, particularly in open-bid auctions. Current suppliers are generally more willing than new suppliers to make dedicated investments toward the buyer. Paradoxically, in sealed-bid auctions, both current and new suppliers increase their willingness to make dedicated investments toward the buyer. Although these auctions can yield cost savings, the savings are category specific and are not systematically related to an open- or sealed-bid format. The author also discusses implications for the use of online reverse auctions in industrial sourcing activities.

In recent years, Internet technologies have yielded a suite of tools to manage sourcing between industrial buyers and their suppliers. These tools aid in contract negotiation, expenditure analysis, supplier selection, risk management, and optimal supplier performance. According to one estimate (Jones 2000), technological efficiency could produce nearly $1 trillion in savings from the $7 trillion annual expenditure on components, suppliers, and services worldwide. How this will happen remains an open question; across virtually all industries, buyers and suppliers wrestle with how to use Internet technology effectively. I am particularly interested in how emerging Internet-based tools will influence the ongoing relationships between buyers and suppliers. Extensive research has already examined successful interorganizational exchanges, but almost none of it considers the interface of emerging technologies with the ongoing relationships between buyers and suppliers. This research addresses that gap.

I consider how the use of online reverse auctions affects a firm's relationship with the supply base. Such auctions have been dubbed "reverse" because sellers bid instead of buyers, and prices are bid down instead of up. These auctions are standard in e-commerce tool kits of companies such as Aruba, Commerce One, Free Markets, Oracle, Bearcats, and various vertical hubs (Scott-Lewis 2001) and have been adopted by leading firms in many industries, including aerospace, automotive, communications, consumer products, pharmaceutical, technology, and the U.S. military (Friedman et al. 2001).

Buyers tend to believe that online reverse auctions are efficient, though anecdotal evidence suggests that the online auction experience can be extremely disturbing: "I am a supplier and very recently had this experience with my major customer. To sit for five hours and watch business that you have developed, maintained, and serviced for 40 years being carved up and slowly disintegrate is a very traumatic experience. Are we seeing the demise of a purchasing staff and sales force, as we know it today?" (posted by Jack Bailey, http://www.purchasing.com, June 15, 2000). A supplier who participated in the online open-bid auctions for this research notes: "[The buyer] talks about the relationship being a partnership and [the auction] really takes that away. There is not a partnership there at all. What they do is take your existing business that you have worked very hard to achieve and maintain. You work with them to give them cost reductions over the years, and they send it out across the board for a competitive bid. I just do not think that is fair." The possibility thus remains that the process may produce grave consequences for supplier attitudes and perceptions of the buyer. This research focuses on this possibility.

I consider online reverse auctions in industrial sourcing contexts and ask questions such as the following: In an online reverse auction, how do open bids versus sealed bids affect a supplier's attitudes and strategic position or bidding performance and motivation? Are suppliers motivated to better serve the buyer? Do the effects differ for current and new suppliers? I investigate how such auctions influence (I) supplier suspicions toward the buyer, ( 2) the buyer's cost savings from the auction, and ( 3) the supplier's willingness to make idiosyncratic investments. The goal is to expose the general effects of online reverse auction processes on supplier attitudes, motivations, and performance. I do not attempt to explain how specific components of these auctions (e.g., number of rounds of bidding, nature of price visibility) affect supplier behavior. Further research can address these questions.

The research involves six quasi experiments in the supply chain of a major industrial manufacturer. Although it is too early to know the long-term effects of online reverse auctions on buyer-supplier relationships, it is possible to examine short-term effects. Drawing on existing economics, marketing, and psychology theories as well as my own experience and extensive interviews with participating sourcing managers, suppliers, and auctioneers, I consider the extent to which these theories generalize to the firm's use of an open- or sealed-bid auction across various products.

This research thus makes several contributions. The effects of Internet technology on relationships between industrial buyers and suppliers can be better understood with a systematic, empirical test in the supply base of a major industrial manufacturer. The research addresses whether interorganizational relations theories are generalizable to online contexts. It also suggests when and how a firm should use auctions for its supply base, which provides insight into ways buyers and suppliers might better organize and manage their electronic procurement activities.

In the next section, I discuss the role of online reverse auctions in the buyer's relationships with suppliers, and I then describe two types of online reverse auctions commonly used in industrial sourcing activities. This is followed by hypotheses as to how supplier attitudes, motivation, and performance might vary between open- or sealed-bid auctions and between new or current suppliers. I describe the research setting and Lest hypotheses. The article concludes with a discussion of results, limitations, and managerial implications.

Relational and Transactional Exchanges

Buyers can develop an array of purchasing arrangements with suppliers. On the one hand, they can take an arm's-length transactional approach, which explicitly defines buyer and supplier roles in preestablished contracts; buying is focused around the specific transaction, and the orientation is win-lose. On the other hand, buyers can foster more collaborative sourcing arrangements that involve specific knowledge and implicit understandings and in which parties focus on mutually beneficial activities and processes. Most purchasing arrangements fall between these two extremes. Heide (1994) further discusses the differences between these two types of exchange.

The present research context involves various sourcing relationship types that range across the spectrum from transactional to collaborative arrangements. Many supply relationships are transactional, yet it is also possible that a supplier's contract is renewed every two to five years on a long-term basis. Consequently, suppliers may have considerable experience with the buyer over the years. To enhance the generalizability of the research, I consider this mix of relationships across six product categories that vary widely in terms of their sourcing and supply management strategies. The only type of relationship not represented in this research is strategic sourcing agreements in which the buyer and supplier are highly interdependent; make mutual idiosyncratic investments; and actively create mutually beneficial, "pie-expanding" opportunities together (see Jap 1999). Suppliers in such relationships are rarely asked to participate in online reverse auctions.

Online reverse auctions have increased in popularity because they emphasize short-term price savings and can simplify and support negotiation. Such auctions have been shown to achieve gross savings (over historical cost in unit prices) from 5% to 40% (Tully 2000), with an average gross savings of 15% to 20% (Cohn 2000). The auctions also drastically reduce the average time involved in negotiation (measured as the point of mailing a request for purchase [RFP] to the compilation of a subset of viable bid offers) from six weeks to a few hours. Online technologies now enable temporal and geographical conveniences, reduced cost of contact, instant feedback, and privacy, which manual auctions cannot offer. Consequently, such auctions are ideally suited for transactional exchange contexts but may be less appropriate for relational exchanges. Some critics have argued that auctions hinder collaboration in relational contexts (Emiliani and Stec 2001); this may be because such auctions do not allow the expression of nonprice attributes, such as quality, service, and reliability. It may also be that such auctions threaten buyers' very existence and purpose in long-term exchanges. Consider the comments of Roger A. Whittier (Purchasing 2001), Director of Corporate Purchasing at Intel: "Intel has run several successful auctions to move surplus equipment and material. [However,] I haven't had that much interest out of buyers. Frankly, when you go to buyers and say we want to start reverse auctions, they feel very threatened by it. They feel they add value as negotiators and through sourcing and so forth. In any kind of business where you actually make a difference by negotiating, picking specifications and having some kind of relationship, then reverse auctions don't make a lot of sense."

Despite this doubt, some supporters maintain that online reverse auctions should play a key role in long-term sourcing arrangements because they provide the greatest payback in a direct purchase expenditure (Scott-Lewis 2001). No systematic research effort has yet examined the impact of online reverse auctions on mutually beneficial relational exchanges between industrial buyers and suppliers.

The Auction Process

I focus on one-sided (one buyer and multiple sellers) sealed-and open-bid auctions, because these two kinds differ starkly in format. Open- and sealed-bid auctions mark opposite ends of a spectrum of auction types that differ in their degree of price visibility (suppliers' ability to view their competitors' bids): Sealed-bid auctions have no price visibility, whereas open-bid auctions have full price visibility for bidders. Most online auction formats are variations of these two types (for a more complete discussion of the evolution and direction of online reverse auctions in industrial sourcing activities, see Emiliani 2000; Jap 2002).

The online reverse auction process typically begins with the buyer posting an RFP to a Web site and inviting specific suppliers to view the RFP. In a sealed-bid auction, suppliers are asked to submit their bids a few days or weeks later, and a winner is subsequently selected. Only the buyer views the bids. In this research, the sealed-bid event involves a single round of bidding. In the open-bid auction, suppliers bid sequentially through a series of product lots or subgroups and can view their competitors' bids and respond in real time. A moving end time (a "soft close") is used for each lot, which means that any bid within the last minute of the closing time automatically extends the end time for a few minutes to allow other bidders to respond.

This section begins by considering the theoretical literature on auctions and then draws on economics, marketing, and psychology research, as well as interviews with auction participants, to develop specific hypotheses. The theoretical literature illuminates the motivational and strategic concerns of participants, and the interviews identify specific organizational characteristics that constitute the sourcing context. A critical set of dependent variables emerges: opportunism suspicions, willingness to make idiosyncratic investments, and cost savings. I consider how these variables may differ across auction (open bid and sealed bid) and supplier (new or current) type.

A Perspective on the Theoretical Literature on Auctions

Many economics studies examine auctions from theoretical and empirical perspectives (for an overview, see Kagel 1995; McAfee and McMillan 1987; Milgrom 1989). Although some similarities exist between the online reverse auctions used in industrial sourcing today and the auctions of the theoretical literature (e.g., the focus on price competition in a structured negotiation format with well-defined rules for the submission and modification of bids), several significant differences make it difficult to generalize from the economics literature to industrial settings. Major differences include product type, winner determination, interdependent bidding practices, and emphasis on bidder behavior over context. First, the products in the auctions of the theoretical literature tend to be commoditized, and price determines products' complete value. In the marketplace, many online reverse auctions may involve products differentiated by price, quality, or other attributes.

Second, the majority of auctions examined in the theoretical literature specify how to determine the winner of the event, typically on a first- or second-price basis (e.g., Bulow and Klemperer 1996; Holt 1980; Milgrom 1989). In the marketplace, however, buyers have full latitude to select the winning supplier on any basis; the only explicit commitment that the buyer makes is to award the contract to one of the participants. Suppliers in the marketplace may not understand how competitive their offer is, but bidders in the auctions of the theoretical literature understand exactly where they stand relative to their competitors and are able to use this information to determine their responses to competitive bids.

A third major difference lies in the sequential bidding of interdependent product lots in online reverse auctions. Bidders must consider not only individual prices but also their capacity to accommodate the lots. This bidding problem can be complex, because the bids placed in the first lot may determine the supplier's bids for the next lot, and so forth. Although some researchers have examined multiunit purchases of homogeneous commodities (Kagel and Levin 2001; Swinkels 2001), interdependence among heterogeneous noncommodities has yet to be addressed.

Finally, the theoretical literature focuses on the processes by which individual actions translate into prices, but it does not focus on the auction context. Economists thus consider differences among bidders' valuations of auction objects, bidder characteristics (e.g., risk aversion), psychological mistakes (e.g., the winner's curse), and the effect on prices in a variety of auction formats (e.g., sealed, open, English, Dutch) that vary in allocation rules. I therefore draw on the economics literature on auctions in a limited way to develop the hypotheses; instead, I rely more heavily on interorganizational theory, particularly transaction cost economics, to determine the influence of online reverse auction processes on supplier attitudes and perceptions of the buyer.

Supplier Suspicions of Opportunism

Although both buyers and suppliers want to make a profit, their approaches differ. Buyers want to reduce the price of purchased materials so as to reduce the cost of goods sold. Suppliers want to maximize sales, particularly through long-term relationships that emphasize quality and delivery. These conditions breed discord and suspicion. The buyer's choice to use a sealed- or open-bid auction may arouse the supplier's suspicion that the buyer is using the auction opportunistically against the supplier.

Opportunism, long a hallmark of the transaction-cost economics framework, is defined as self-interest-seeking with guile. It is synonymous with misrepresentation, cheating, and deception and subsumes a range of misbehavior, such as adverse selection, moral hazard, shirking, subgoal pursuit, agency costs, and free riding (Williamson 1996); it also has received growing attention in recent years (Brown, Dev, and Lee 2000; Wathne and Heide 2000). It is worth noting that opportunism is not merely a form of distrust. Trust is a broad metaconstruct with many facets and levels; scholars across multiple disciplines do not fundamentally agree on the meaning of trust (Rousseau et al. 1998). Opportunism is more delimited and behavioral in nature; it is observable by the supplier and grounded in specific actions and should create reduced attributions of trust.

I focus on the supplier's suspicions (i.e., its perceptions) that the buyer is acting opportunistically, rather than on proven opportunism, because the supplier typically cannot verify the buyer's guile. The supplier's behavior is affected by its misgivings about the buyer's character (cf. Rusbult and Van Lange 1996); when the supplier suspects the buyer, the supplier usually holds back from the relationship to avoid vulnerability to further opportunism (Ping and Dwyer 1992; Williamson 1985, 1993). Suspicion might also motivate the supplier to use additional safeguards (contracts, incentives, or monitoring) to protect and enforce the exchange.

Sealed- versus open-bid online reverse auctions. Which auction format raises supplier suspicions of opportunism in industrial contexts? I expect that the supplier's opportunism suspicions increase in online open-bid reverse auctions more than in sealed-bid reverse auctions, because price competition is greater and more explicit in open-bid auctions. The fast-paced, dynamic bidding, along with the need to respond quickly to competitors' bids, yields tense negotiation and pressure on suppliers to cut prices vigorously. An open-bid auction increases the supplier's bargaining costs, which makes the process so disagreeable to suppliers that they will accept renegotiation rather than persist with current pricing levels (see Masten 1988). For the supplier, the open-bid format can force additional price concessions from the supplier, becoming a form of opportunistic rent seeking on the part of the buyer.

It could be argued, however, that price visibility in open-bid auctions should reduce supplier suspicions, because suppliers can gauge their bids relative to the competition and can choose whether to reveal their own value of the contract by responding to competitive bids. In contrast, the optimal bidding strategy in a sealed-bid auction demands that suppliers reveal their bottom-line bids, because there is no opportunity to change the bid, view competitors' bids, or respond to others' bids. In this sense, sealed-bid auctions appear more opportunistic.

In field interviews with suppliers, I learned that the compressed time frame of open-bid auctions creates a stressful context for the supplier. In private conversations, many suppliers complained that the format prevented them from carefully considering price bids and gave them a sense of being "out of control."

H₁: The increase in supplier opportunism suspicions before and after the auction is greater in online reverse open-bid auctions than in sealed-bid auctions.

New versus current suppliers. Opportunism suspicions may also differentially increase between new and current suppliers. Current suppliers may have a rich history of exchange with the buyer. They understand the buyer's needs and constraints and may benefit from trust, implicit understandings, or relational norms. Their history could act as a switching cost, making it difficult for the buyer to choose a new supplier. But new suppliers do not have the benefit of past experience, which significantly reduces their bargaining power. A buyer's decision to use an electronic format could increase a current supplier's suspicions.

H₂: In online reverse auctions, the increase in supplier opportunism suspicions before and after the auction is greater for current than for new suppliers.

Supplier Willingness to Make Idiosyncratic Investments

The dedicated investments made by buyers and/or sellers are a key aspect of the relationship. Such investments may be tangible (e.g., plant equipment, tooling, design systems) or intangible (e.g., human resources, training) and typically increase the effectiveness and efficiency of one or both parties (Dyer and Singh 1998; Heide 1994; Lusch and Brown 1996; Noordeweir, John, and Nevin 1990). The supplier's idiosyncratic investments are a risk on behalf of the buyer to produce superior returns and joint value; they represent a supplier's commitment to the buyer (Anderson and Weitz 1992) and are difficult for the supplier to redeploy if the relationship ends.

Sealed- versus open-bid auctions. I expect that the supplier's willingness to make idiosyncratic investments decreases in open-bid auctions, because the format's rapid, dynamic price competition emphasizes price reduction more strongly than in sealed-bid formats. Open-bid auctions may also signal a focus on price in the short term, which fosters a market governance structure that focuses on discrete transactions. Such a relationship does not foster mutual value or benefit in the long run. The supplier's incentive to make idiosyncratic investments is therefore reduced, because payback is unlikely to be realized over time.

H₃: The decrease in a supplier's willingness to make idiosyncratic investments before and after the auction is greater in online reverse open-bid auctions than in sealed-bid auctions.

The opposite effect can be argued. Availability of information about pricing in an open-bid auction clarifies the supplier's standing relative to the competition and provides additional information about the likelihood of recouping the value of dedicated investments in the exchange. After observing competitive pricing, suppliers may be motivated to make idiosyncratic investments to attain superior value, improved coordination, or scale economies. In contrast, a sealed-bid auction does not provide feedback about a supplier's relative standing but rather encourages the supplier to minimize its bid price at all costs.

New versus current suppliers. As buyers turn to online auctions in sourcing, the motivation for current suppliers to make idiosyncratic investments may be reduced, because the current suppliers' value--their knowledge of the buyer and the buyer's specific needs--is not readily conveyed in electronic contexts. Little opportunity exists to communicate and explore potential mutually beneficial activities and processes stemming from idiosyncratic investments. As negotiation focuses on short-term price, suppliers have little incentive to create long-term investments with the buyer; accordingly, I anticipate the following:

H₄: In online reverse auctions, the decrease in willingness to make idiosyncratic investments before and after the auction is greater for current suppliers than for new suppliers.

Buyer Cost Savings

A chief concern of the buyer is cost savings, which is defined as the percentage reduction in price from historical costs, and it is an important metric in industrial procurement. Does a sealed- or an open-bid auction yield higher savings? There is a significant literature in economics that considers this question for auctions with well-defined parameters (e.g., specific allocation rules, revealed number of bidders, single commodities). These results indicate that when the bidders have common values, the open-bid auction format produces greater cost savings (Milgrom and Weber 1982). When bidders' values for the contract are common, the value of the item is the same to all bidders, but bidders have different information about the underlying value. These characteristics are true of industrial sourcing auctions. A true value exists for the contract--the worth of the contract in the market--but no one knows the true value, and each bidder estimates differently as to how much the item is objectively worth. In an open-bid format, the bids partially make public each bidder's private information about the true value of the contract. Each bidder is thus able to learn from the bidding process and adjust its bid closer to the true value of the contract.

It is important to note that these results hold in auctions in which the rules of allocation are clear; in this research, the rules of allocation are more ambiguous. The buyer retains full latitude in selecting the winner, regardless of the nature of the price bids, which may influence bidders' motivation such that they do not bid as predicted by economic theory. Whether the theory is able to generalize to the open- and sealed-bid auctions of the current marketplace remains an open question. Thus, the hypothesis from the theoretical literature is as follows:

H₅: The buyer's cost savings are greater in online reverse open-bid auctions than in sealed-bid auctions.

New versus current suppliers. Do new or current suppliers provide greater savings? The auction literature does not address this issue. I believe the question is empirical and yields three possible outcomes. First, it could be argued that new suppliers should be more aggressive than current suppliers, because new suppliers gain not only a purchase contract but also the potential to remain in the supply base for the long run. Second, it could be argued that current suppliers should be more aggressive than new suppliers, because the current suppliers have much to retain. These suppliers have built a history of exchange and learning with the buyer. A third possibility is that both effects operate simultaneously and produce no discernible empirical differences. I reserve prediction on the direction of this result and examine the empirical outcome in a subsequent section.

Research Setting

I conducted the research from 1999 to 2000 in the supply base of a major firm in the automotive industry that bought various components and parts. The automotive industry was an early adopter of online reverse auctions, having developed Covisint, one of the largest online markets in which online reverse auctions play a significant role in materials sourcing for major automobile manufacturers. In 2001, DaimlerChrysler held a reverse auction through Covisint, and the total order volume exceeded $2.5 billion, which made it the largest, single, Internet-based auction. The participant firm in this study rarely engaged with suppliers in long-term strategic partnerships. Instead, the buyer's mix of supplier relationships varied in product categories along the spectrum from transactional to collaborative sourcing arrangements.

In the fall of 1999, the company experimented with various types of auctions over a six-month period and evaluated the financial, relational, and strategic consequences. Sourcing managers were asked to host either an open-bid or a sealed-bid online auction. As an independent researcher, I could make recommendations to the firm but ultimately could not intervene as much as I would have liked; for example, I was unable to control the number of suppliers, lots, or product types in each event. I was only allowed to survey suppliers before and after the event, to conduct postauction interviews with suppliers, and to interview sourcing managers throughout the process.

Approximately $100 million in purchase contracts was made available for bid in three open-bid and three sealed-bid auction events in six different product categories. None of these products were pure commodities, such as maintenance, repair, and operations supplies or highly customized strategic parts. All the products were used in production or directly for parts in production. The product categories thus differed in nonprice characteristics, and supplier relationships could play a significant role in exchange. Table I provides an overview of the product categories and the number of bidders and lots in each auction.

The sourcing manager qualified (through visits, questionnaires, and research) a list of viable suppliers before each event and then invited a subset of suppliers to hid in the auction. All the auctions were conducted by a third-party auctioneer who informed the suppliers of the event rules: (I) The sourcing manager committed to select a winner from each event on any basis, but the lowest bid was not guaranteed to win the contract; ( 2) the sourcing manager was prohibited from bidding against suppliers in the auction (an unethical practice known as "shilling"); ( 3) all competitors were viable prequalified sourcing options for the buyer; and ( 4) supplier bids were legally binding. The suppliers were not told who their competitors were or how many suppliers would bid against them.

Because suppliers in the open-bid auctions had virtually no experience with the format, the auctioneer trained them to use the online interface before the auction so that they could understand cost structures and bid at, not below, their marginal costs. The auctioneer also helped the supplier imagine various scenarios to help mitigate the pressure of the real-time bidding decisions. The auctioneer's role worked against the probability of my finding significant results. By encouraging suppliers to provide high-quality, educated, and enlightened bids, the auctioneer decreases the likelihood that observed changes in the dependent variables of interest are due to confusion, misunderstanding, or other problems that might arise from suppliers being ill-prepared for the open-bidding experience.

Data Collection

The data collection was a multimethod approach that involved extensive interviews with participants and an untreated-control-group quasi experiment with a pretest and posttest around the auction events (Cook and Campbell 1979). The treatment variables were auction type (i.e., scaled/open) and supplier type (i.e., new/current), and the dependent variables were suspicions of opportunism, willingness to make idiosyncratic investments, and cost savings. The field design featured an array of controls, including ( 1) pretest and posttest survey measures from the same panel of suppliers in the treatment group. ( 2) control groups, and ( 3) replication of the design across six product categories. I initially tried to administer the control group survey twice, to correspond to the pretest and posttest surveys, but found that no noticeable change occurred over a week. In addition, suppliers resisted completing the survey twice with little change of their circumstances in between.

Procedure. In each product category, in-depth interviews were conducted with sourcing managers to obtain a better understanding of the exchange context, the composition of the supply base, and the expectations and strategic intentions for each auction. Managers provided the names and contact information of the qualified suppliers in the treatment group. Because buyers do not always involve their entire supply base in a negotiation, for the purpose of this research, the sourcing manager provided contact information for suppliers not invited to the event but equivalent to the treatment group (e.g., in production capability, price competitiveness, and product offerings). These suppliers formed the control group. The managers understood the purpose of the control group within the research design and carefully selected participants.

Suppliers in the treatment and control groups were sent an e-mail invitation one week before the event that specified that the respondent should be knowledgeable about the firm's specific relationship with the buyer. For the treatment condition, the invitation also asked that the respondent be someone who would participate in the upcoming bid process. Suppliers in the control group were not told that the buyer was hosting a competitive bid event in their product space; they were merely asked to participate in a study on buyer-supplier relationships. The invitation directed suppliers to the survey on a university Web site; it also guaranteed anonymity and reassured suppliers that the buyer would not have access to individual responses. The survey directed the supplier to complete all items in reference to the specific buying organization. One week after the auction, suppliers in the treatment group were sent an e-mail invitation to complete the posttest survey. Throughout the data collection, I monitored the buyer's activities to ensure that no major events or initiatives (e.g., retroactive charge backs) occurred to disrupt or alter supplier perceptions and attitudes.

Respondents. Respondents were typically senior executives, vice presidents, or owners of a supply business who handled large customer accounts and had the authority to determine major investment decisions and make price concessions. At the time the research was conducted, online auctions and electronic bidding had hardly permeated the marketplace, such that suppliers had virtually no experience with the formats.

Of the 154 bidders, 68 completed the surveys, which yielded a response rate of 44%. 01 the 68 bidders, 33 responded in the open-bid auctions, and 35 responded in the sealed-bid auctions. Twenty-eight of the 68 respondents were new suppliers and 40 were current suppliers. The control group was composed of 87 suppliers; 50 of these suppliers completed the survey, which yielded a response rate of 57%. Of the suppliers in both the treatment and the control groups, 33% were new suppliers and the remainder were current suppliers.

Respondent competency. Because some of the hypotheses rely completely on suppliers' perceptions, the respondents must be competent to report, and differences in respondent knowledge, position, and perceptions should be minimized. This end is accomplished by global and specific measures of respondent competency and knowledge of the phenomena. The global measure was the respondent's tenure with a firm. The respondents averaged 6.3 (standard deviation [s.d.] = 5.9) years of experience in their area and had been with their firms 103 (s.d. = 5.1) years on average.

The respondent's knowledge of the relationship with the buyer was assessed by questions at the conclusion of the pretest and the control questionnaire. The respondents were asked, "How knowledgeable are you about the following aspects?" Listed below were items such as "Your firm's willingness to work with [the buyer firm]," "The degree to which your firm trusts [the buyer firm]," and "Your firm's willingness to invest in a customer," Responses varied along a seven-point rating scale (1 = "not very knowledgeable" and 7 = "very knowledgeable"). Throughout this article, all scale scores of multi-item measures reflect the mean score of the multiple items, not a sum score. The average response to these items was 6.4 (s.d. = .57). Collectively, some confidence emerged that the selected respondents were knowledgeable, relatively involved in the survey, and unlikely to have fabricated answers to the items.

Measurement. All the scales used in this study are listed in the Appendix. The supplier's opportunism suspicions refer to bad faith in the buyer. To measure bad faith and describe it in a survey format, I conducted multiple prestudy interviews with sourcing managers and suppliers in other industrial product categories. On the basis of these interviews, I decided to measure supplier suspicion with a four-item scale that reflects the firm's specific opportunistic behaviors (e.g. reneging, lying, falsifying information) as perceived by the supplier. The supplier's suspicion of opportunism shows a mean of 2.94 (s.d. = 1.40) that ranges in value from 1 to 7; Cronbach's alpha coefficient for this scale is .79.

The mean value reflects the relationship of a powerful buyer and a lower-tier supplier. In the automotive industry, higher-tier buyers regularly exercise power over suppliers. It is thus natural that the suppliers would suspect the buyer of opportunistic behavior, though theoretical reasons also help explain this result. Williamson (1993) notes that firms tend to engage in business as usual rather than opportunism, partly because managers are well socialized and partly because governance structures mitigate opportunism. If opportunism is too high, the supplier would likely not exchange with the buyer at all.

The supplier's willingness to make idiosyncratic investments is a modified version of the scale developed by Cannon (1992) that reflects tangible and intangible investments that the supplier would find difficult to move into another relationship. The four-item scale indicates the extent to which the supplier is willing to make investments in training, production procedures, equipment, tools, and capacity to accommodate the buyer. The supplier's willingness to make idiosyncratic investments shows a mean of 4.50 (s.d. = 1.68) that ranges in value from I to 7; Cronbach's alpha coefficient for this scale is .87. The correlation between this construct and opportunism suspicions is .12 (p < .17).

I assessed the measurement properties of these two scales using a confirmatory factor analysis with maximum-likelihood estimation methods in LISREL 8.03 (Jöreskog and Sörbom 1993). The estimated measurement model of the two latent factors, each with four reflective indicators, has a chi-square fit of 38.93 with 19 degrees of freedom (p < .00). The comparative fit index is .92, the incremental fit index is .93, and the Tucker-Lewis fit index is .89. All the factor loadings and measurement errors are in acceptable ranges and are significant at α = .05, which provides evidence of convergent validity. The pair of constructs also passes the test of discriminant validity that Fornell and Larcker (1981) recommend.

Analysis of Hypotheses

To assess the impact of the open- and sealed-bid formats on supplier attitudes, I considered how the supplier scores for these measures changed from the pretest to the posttest. Because the quasi experiment is a nonequivalent design (due to its nonrandom selection process), the expected values of at least one characteristic of the groups are not equal even in the absence of a treatment effect (Cook and Campbell 1979). To obtain a reasonable estimate of the treatment effect, the analysis therefore must properly recognize and account for the effects of these initial differences; two steps accomplish this purpose. In the first step, the respondents in the pretest and control group were compared in terms of their demographics and intangible aspects along several dimensions, such as annual sales, willingness to collaborate, satisfaction with the relationship, and perceived dependence on the buyer. No significant differences were found. For reasons of space, the specific means and t-tests are not presented; the scales are listed in the Appendix.

Although the lack of differences suggests that the groups are equivalent, the groups might have shown large expected differences on other variables that could have affected the posttest scores. These differences were accounted for in the second step of the analysis, in which I used a between-subjects nested design analysis to examine the hypotheses in the treatment group. The auction type was specified as a fixed factor, and the product category and supplier responses were specified as nested random factors. From these factors, the following equation, in which I used the pretest measure as a covariate to adjust for initial differences between groups, is estimated:

( 1) Posttest = treatment + product_(treatment) + pretest,

where

posttest = the posttest score on the variable of interest,

treatment = the treatment effect for open/sealed auctions or new/current suppliers,

product_(treatment) = the effect of product differences nested within each treatment effect, and

pretest = the pretest score on the variable of interest.

This analysis of covariance (ANCOVA) matches the pretest value to the predicted posttest value for all levels of the treatment and examines the differences between them. A statistically significant treatment effect would thus suggest that one level of the treatment would have significantly outperformed the other level, thus controlling for differences in the pretest scores. By matching the pretest measure to each posttest measure, the ANCOVA has more power to detect true differences in the treatment effect than an elementary analysis of variance, which would consider only the posttest responses. I also assessed the extent to which differences exist across product categories (differences in the value of the purchase contract, number of bidders, response rates, and number of lots). A significant effect of product_(treatment) indicates a significant difference among the product events nested within auction type.

At this point, some readers might be concerned that suppliers would always suspect buyers of being opportunistic, expressing negative attitudes toward suppliers, or overstating their willingness to make investments. Such biases could increase the scores that I observed. Even so, such biases would be operative in both the pretest and the posttest but would not explain a change in the scores over time, which is the focus of this research. It is also worth noting that at the time of the posttest, suppliers had not yet been informed of whether they had won the auction, which reduces the likelihood that their responses to the posttest measures are simply a reflection of being "sore losers."

Examination of H₁. H₁ posits that the increase in opportunism suspicions is greater in open-bid than in sealed-bid auctions. In the open-bid auction, the pretest mean was 2.64 (s.d. = 1.27), and the posttest mean was 4.29 (s.d. = 1,17); the difference is significant (t₆₄ = 5.51; p < .00). Throughout this article, all t-tests are one-tailed. The pretest mean in the sealed-bid auction was 2.38 (s.d. = 1.02), and the posttest mean was 2.84 (s.d. = .99); the difference is marginal (t₆₈ = 1.94; p < .10). The ANCOVA has an explained variance of .40. A statistical test of the treatment effect of auction type indicates that the effect is significant (F = 16.21; p < .05), which suggests that the increase in scores is significantly greater in the open-bid auction than in the sealed-bid auction. I found no significant differences (F = 1.35, not significant [n.s.]) across the various product categories. Together, the results suggest that opportunism suspicions do not change in sealed-bid auctions but increase in open-bid auctions, and the difference between the conditions is significant, in support of H₁.

Does a self-selection bias explain the increase in suspicions? I considered the possibility that suppliers who responded to the posttest in the open-bid condition were potential "sore losers" (did not offer high savings) of the auction and therefore might be more likely to express opportunism suspicions. This possibility would mean that suppliers who believed they had bid poorly in the auction would be more likely to express their negative attitudes toward the buying organization in the posttest survey than would suppliers who did not bid poorly (offered high savings). In the case of such a bias, it should be observed that suppliers who responded to the posttest offered systematically lower cost savings than did the suppliers at the pretest. However, a t-test of the savings provided by suppliers who responded to the pretest and posttest survey revealed no significant differences (t₆₄ = -.78; p < .44), which suggests that it is unlikely that a self-selection bias accounts for the results.

Does the auction process explain the increase in opportunism suspicions? It was proposed that suppliers view the open-bid process as an unfair means by which the buyer gains concessions from the supplier. This possibility is further explored in several control items included in the posttest survey to capture the supplier's direct view of the auction process. Suppliers indicated their response to these items using a scale of 1 = "strongly disagree" and 7 = "strongly agree." One item stated: "This process will reduce my chances of earning a fair margin on the business." The suppliers in the open-bid auctions averaged a response of 48 (s.d. = 1.4), and suppliers in the sealed-bid auction showed a mean response of 3.0 (sd. = 1.4). The difference between the means is significant at α = .05. Another item stated: "This process does not give a supplier a fair opportunity to bid on business." Suppliers in the open-bid auction showed a mean response of 4.1 (s.d. = 1.1), and suppliers in the sealed-bid event showed a mean response of 2.7 (s.d. = 1.6). Again, the difference in means is significant at α = .05. Both of these items reflect the sense of exploitation suppliers associate with auctions Together, they suggest that suppliers in the open-hid auctions consider the auction process more opportunistic than do suppliers in the sealed-bid process.

Additional indicators of suspicions of opportunism. After the auctions, t conducted interviews with the suppliers in the open-bid condition to assess the difference in opportunism suspicions across the treatment conditions These interviews were not intended to provide causal evidence but rather to explain the difference In the open-bid condition, the suppliers' responses show that they believe the buyer purposely selected an unfair price-competition structure. Three themes emerge The first is that suppliers believe the buyer is using the open-bid format to survey the market pricing without any intention of awarding the business As one supplier said, "All they were going to do was just feel out what the numbers were going to be. Let's say they're looking at someone in Brazil or some Korean firm out there. At this point in the junction, they weren't going to go with those guys based on what they were. So all these [people] were throwing low bids, but it had no meaning as to what was going on." The second theme indicates that the supplier believes the buyer has created false competition by including nonviable bidders, as in this example: "I didn't think the competition I was dealing with in the atmosphere I was quoting really had the wherewithal that I had to supply the parts and do the things I had to do." The third theme is the supplier's belief that the buyer is shilling its bids to push down the price artificially.

The irony is that none of the supplier perceptions correspond to reality. As an independent observer, I knew that buyers had every intention of selecting a winner and had invited only viable suppliers to bid on the purchase contract. No shilling occurred; the buyer and the auctioneer went to great pains to avoid such perceptions by clearly communicating the rules of the game. Thus, a significant gap exists between supplier perceptions of the event and reality, which should be cause for concern.

Examination of H₂. H₂ considers whether opportunism suspicions increase more for current than for new suppliers. The pretest mean of new suppliers was 2.36 (s.d. = 1.30), and the posttest mean was 4.19 (s.d. = 1.38); the difference between the two means is significant (t₅₄ = 5.11; p < .00). For current suppliers, the pretest mean was 261 (s.d. = 1.03), and the posttest mean was 3.10 (s.d. = 1.03); the difference in these means is also significant (t₇₈ = 2.11; p < .05) The ANCOVA has an explained variance of .48. A statistical test indicates that the overall effect between new and current suppliers is not significant (F = 2.19, n.s.), and the differences among product categories are not significant (F = 2.60, ms.) Consequently, although opportunism suspicions increase, this increase does not differ between new and current suppliers.

Examination of H₃. H₃ considers whether the supplier's willingness to make idiosyncratic investments decreases more in open-bid auctions than in sealed-bid auctions. In open-bid auctions, the pretest mean was 4.89 (s.d. = 1.34), and the posttest mean was 5.32 (s.d. = .90); these means do not differ significantly (t₆₄ = 1.48; n.s.). In sealed-bid auctions, the pretest mean of the supplier's willingness to make investments was 4.13 (s.d. = 1.67) and the posttest mean was 4.86 (s.d. = 1.23); this difference is significant (t₆₈ = 2.10; p < .05). The ANCOVA has an explained variance of .32. A statistical test of the treatment effect of auction type indicates that the effect is not significant (F = 1.95, n.s.), which suggests that the mean increase does not differ across auction type, nor does any significant (F = .03, n.s.) effect of differences in means exist across the product categories. Collectively, these results suggest that in the sealed-bid condition, suppliers increase their willingness to make idiosyncratic investments whereas suppliers in the open-bid condition do not change.

Why does willingness to make investments increase in the sealed-bid condition? Posttest interviews provide some speculative answers to this question. In the sealed-bid auctions, suppliers view the process as an improvement over the manual processes of the past. They interpret the electronic bidding process as a signal that the buyer is informed about recent technological developments, which suggests a mutual orientation and potential benefit for both exchange parties. A buyer's technological stance toward mutual benefit increases the supplier's willingness to make dedicated investments.

Examination of H₄. H₄ considers whether the decrease in willingness to make idiosyncratic investments is greater for current than for new suppliers. For current suppliers, the pretest mean was 4.03 (s.d. = 1.64), and the posttest mean was 4.96 (s.d. = 1.18); the difference is significant (t₇₈ = 2.91; p < .00). The pretest mean of new suppliers was 5.18 (s.d. = 1.18), and the posttest mean was 5.26 (s.d. = .95); the difference is not significant (t₅₄ = .30; n.s.). The ANCOVA has an explained variance of .21. A statistical test of the treatment effect of supplier type is not significant (F = 1.16, n.s.), and the differences across product categories are also not significant (F = 1.19, n.s.). These results indicate that current suppliers increase their willingness to make idiosyncratic investments to match the level of new suppliers such that no significant difference exists after the auction in willingness across supplier type.

Examination of H₅. H₅ examines whether the difference in cost savings is greater in open-bid than in sealed-bid auctions. Because this information is extremely sensitive to the host firm, I do not reveal the absolute magnitude here. Instead, I provide standardized indicators to give readers a sense of the relative magnitude of the savings across auction type. The unit of analysis for testing this hypothesis is the lowest bid of each lot. This number is then divided by the historical price and subtracted from unity to generate a cost-savings percentage from the auction. Because the supplier bids are not normally distributed, I used the logarithms of the cost-savings percentage offered in each lot to make the data less nonnormal. This logarithm is then averaged with the savings of the other lots within the same auction type to produce a mean savings for the auction type. As a result, 21 lots produced varying cost-savings percentages in the open-bid auctions, and I averaged these together to produce the average cost savings in the open-bid auctions (X_O); similarly, I computed a mean for the sealed-bid condition (X_S). I indexed the mean of these two numbers (X_OS) to 100. The mean sealed-bid auction savings can then be reflected as 92, and the open-bid auction savings can be reflected as 108. The ANCOVA has an explained variance of .16. A test of the treatment effect indicates that it is insignificant (F = 1.29, n.s.); however, a statistical test of the effect across product categories is significant (F = 13.28; p < .00). Collectively, these results indicate that though the savings across auction types does not differ, significant variation exists across product categories.

I also examined the differences in savings offered by new and current suppliers. In this case, the unit of analysis is the individual supplier's bid. I averaged the cost-savings percentage offered by new suppliers and the percentage offered by current suppliers; they differ by less than one percentage point. The ANCOVA has an explained variance of .18. A statistical test of the treatment effect indicates that it is insignificant (F = .53, n.s.); however, the effect of product categories is significant (F = 5.75; p < .001). These results indicate that differences in savings occur because of differences across product categories and do not vary systematically across auction types.

The results of the quasi experiment indicate that online reverse auctions increase both new and current suppliers' beliefs that the buyer acts opportunistically toward the supplier, particularly in open-bid auctions. Paradoxically, the supplier responds to online auctions by increasing its willingness to make dedicated investments when the buyer uses a sealed-bid auction format. The results also show that current suppliers increase their willingness to make idiosyncratic investments toward the buyer. Although online reverse auctions can yield cost savings, this savings is category specific and is not systematically related to an open- or sealed-bid format. Together, these results indicate that online reverse auctions can exert complex relational effects on the supply base.

Supplier Suspicions of Opportunism

The results show that suppliers' suspicions of opportunism increase after an open-bid auction. Suppliers consider the open-bid process exploitative and unfair. After the event, suppliers voiced their resolution to avoid such events and condemned the process: "In the future, I would never play this game again. We'll play it, but what will happen is we'll be even more adamant as far as to what our prices are. It wasn't a very professional way to handle the business." The supplier's inference of the buyer's opportunism is critical, because it affects suppliers' view of exchange governance and their subsequent actions. In the future, suppliers might demand more explicit contractual assurances or contingency agreements to safeguard their returns. To maintain the lower pricing scheme, they might be forced to reduce quality, value-added services, or overall responsiveness to the buyer, all of which are features that might also be withdrawn to retaliate against the buyer. Consequently, the supplier's suspicion of opportunism might motivate it to respond in kind (Axelrod 1984).

Suppliers in the automotive industry are already beginning to organize themselves against online reverse auctions (Kisiel 2002), claiming that buyers abuse the auctions and calling for the creation of a formal code of conduct to discourage what they perceive as opportunism. These suppliers contend that online reverse auctions have far outnumbered the actual contracts awarded and that buyers use these auctions to determine how low suppliers are willing to bid and then wring additional price concessions from current suppliers. This scenario illustrates how the mere perception of opportunism surrounding these online reverse auctions can poison relationships between buyers and suppliers.

The results also indicate that opportunism suspicions are heightened for both new and current suppliers, contrary to my predictions. For current suppliers, online reverse auctions may inhibit their ability to sell intangible aspects to the buyer. More than 50% of the suppliers in the postevent interviews considered this inability to express their full capabilities a disadvantage to electronic bidding mechanisms. It is notable that new suppliers also display heightened suspicions. They may consider electronic bidding a poor medium for sales, or perhaps they consider such suspicions a cost of doing business with a powerful supplier. Although the evidence and possibilities are intriguing, further research must determine the appropriate explanation.

Supplier Willingness to Make Idiosyncratic Investments

One notable result of this research is suppliers' increased willingness in the sealed-bid condition to make idiosyncratic investments in transactional exchanges. Suppliers evidently consider the buyer's choice of an online sealed-bid auction an attempt to improve the transaction, which is taken as a signal of mutual orientation and encourages suppliers to make dedicated investments. This, in turn, increases the likelihood that suppliers will recoup the value of their investments in the long run.

The data also suggest that online reverse auctions may provide a useful wake-up call to current suppliers, which are as willing to make idiosyncratic investments in the buyer as are new suppliers. Current suppliers may also be signaling their commitment to the buyer in the long run.

Buyer Cost Savings

Another result of this research is that cost savings do not differ systematically across the online reverse auction type or supplier type but may be affected by different characteristics and conditions of the auction event. Perhaps suppliers have not yet developed a long-term strategy to handle such auctions. Because this research involves only six online reverse auctions, I cannot investigate more fully how differences in the number of bidders, lots, size of purchase contracts, or other characteristics might have been systematically related to the level of savings for each event. The decision to use online reverse auctions for direct sourcing activities is complex, and its success depends on various conditions yet to be identified and understood.

Collectively, the results paint an intriguing picture of supplier reactions and financial implications in transactional exchanges. The prevailing theory predicts that economic actors are shortsighted, self-interested profit maximizers that respond to the signals and behaviors of their exchange partners. In reality, firms are not as shortsighted and non-strategic as might be believed; rather, they make complex decisions about their exchanges. Perhaps this strategy is rational for firms that must make both short- and long-term Choices.

Limitations

The research exhibits some limitations. It is specific in scope because it considers only two types of one-sided online reverse auctions in transactional exchanges. The observed effects may not generalize to alternative online reverse auction formats, to procurement of indirect materials, or to close partnerships. In addition, the supplier training may have decreased the amount of variation that was observed in the bids. It is also not clear whether the observed effects will last, because measurements were taken after only one online reverse auction. However, the extent to which I witnessed any noticeable changes in supplier attitudes and perceptions is notable, given that these constructs are typically quite stable over time. The online reverse auction thus exerts a fairly significant impact on suppliers in the short run. Because this was a field experiment, I had limited discretion as to its design and level of control. This research is not a pure test of the theoretical literature on auction theory; rather, it has the advantage over laboratory experiments of examining experienced professionals with larger, real sums of money at stake.

Implications for Management

The results demonstrate that open-bid, online reverse auctions can raise supplier suspicions of buyer opportunism. Buyers should therefore be selective in their use of these auctions, perhaps limiting them to purchases involving less important supplier relationships, such as the purchase of indirect materials. Because the purchase of indirect materials can constitute anywhere from 40% to 60% of a firm's total purchasing volume, a sizable opportunity remains for auctions to exert a significant impact on sourcing costs.

Online reverse auctions could also be used as a screening mechanism for long-term sourcing arrangements, which would capitalize on a current supplier's willingness to make the necessary investments, to assure the buyer that it receives competitive pricing, and to mitigate opportunism suspicions in the long run. Some firms already use online reverse auctions in this way. Emily Andren of the Gartner Group notes that, "Initially, most reverse auctions were used for spot buying, but companies are increasingly using them to select suppliers for long-term contracts."

Another implication is that the supply base may not be able to provide constant price reductions over the long run. Over time, suppliers may need to leave the industry because of their inability to compete, or they may consolidate to reach the scale economies to support lower prices. Both possibilities reduce the number of alternative suppliers for the buyer and shift power to suppliers. Buyers should therefore carefully consider using online reverse auctions repeatedly in the long run.

Directions for Further Research

Further research might consider whether online reverse auction formats that reveal less information, perhaps only the lowest market bid or a rank ordering of the bids, would produce different effects from what this research indicates. More work is needed on the circumstances that create cost savings. Do specific supplier characteristics, conditions in the supply base, or buyer-sourcing strategies yield significant savings? What role do the numbers of bidders, lots, rounds of bidding, and size of the purchase contract play in motivating how suppliers bid in various types of online reverse auctions? Further research should consider these questions across many more auctions.

The results also raise the question of how online reverse auctions affect supplier behavior over time. Do buyers observe a reduction in quality and service? What other effects exist on supplier motivation and attitudes toward the buyer? Experimental evidence from common value auctions suggests that "market learning" occurs as bankruptcies drive out the aggressive bidders and as more aggressive bidders earn lower than average profits. Garvin and Kagel (1994) observe that over time bidders begin to self-select out of future auctions. They also respond to repeated losses by bidding less. Further inquiry must determine whether this result generalizes to the marketplace.

Finally, this research is the first field experiment in the marketing literature on interorganizational relations. Although such quasi experiments are plentiful in other areas of marketing (e.g., advertising), interorganizational researchers have not used them. The firm's need to leverage Internet technologies offers a prime opportunity for researchers to find and test unique predictions and to conduct longitudinal tests and quasi experiments that will better enhance the understanding of the role and value of emerging technologies in marketing strategy.

Legend for Chart:

A - Auction
B - Product
C - Number of Bidders
D - Number of Lots

  A                  B                       C         D

Sealed       Transportation                  72        81
             Nonproduction services           8         2
             Semiconductors                   7         3

Open         Plastics                        12         1
             Electrical parts                35         5
             Metal parts                     20         6

Totals                                      154       105

Anderson, Erin and Barton A. Weitz (1992), "The Use of Pledges to Build and Sustain Commitment in Distribution Channels," Journal of Marketing Research, 29 (February). 18-34.

Axelrod, Robert (1984), The Evolution of Cooperation. New York, Basic Books.

Brown, James R. (1981), "A Cross-Channel Comparison of Supplier-Retailer Relations," Journal of Retailing, 57 (Winter), 3-18.

-----, Chekitan S. Dev, and Dong-Jin Lee (2000), "Managing Marketing Channel Opportunism: The Efficacy of Alternative Governance Mechanisms," Journal of Marketing. 64 (2), 51-65.

Bulow, Jeremy and Paul Klemperer (1996). "Auctions Versus Negotiations." The American Economic Review, 86 (1), 180-94.

Cannon, Joseph P. (1992), "A Taxonomy of Buyer-Seller Relationships in Business Markets," doctoral dissertation, Department of Marketing, University of North Carolina, Chapel Hill.

Cohn, L. (2000), "B2B: The Hottest Net Bet Yet?" BusinessWeek, (January 17), 36-37.

Cook, Thomas D. and Donald T. Campbell (1979), Quasi-Experimentation: Design and Analysis for Field Settings. Boston: Houghton Mifflin.

Dyer, Jeffrey H. and Harbir Singh (1998), "The Relational View: Cooperative Strategy and Sources of Interorganizational Competitive Advantage," Academy of Management Review, 23 (4), 660-79.

Emiliani, Mario L. (2000), "Business-to-Business Online Auctions: Key Issues for Purchasing Process Improvement," Supply Chain Management, 5 (4), 176-86.

----- and David J. Stec (2001), "Squaring Online Reverse Auctions with the Caux Round Table Principles for Business," Supply Chain Management, 6 (3), 101-105.

Fornell, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error: Algebra and Statistics," Journal of Marketing Research, 18 (August), 382-88.

Friedman, Jamie, Thomas P. Berquist, Chris Debiase, Steven Kahl, and Cheng Lim (2001), "Technology: B2B Software," research report, Goldman Sachs. (February 23).

Garvin, Susan and John H. Kagel (1994), "Learning in Common Value Auctions: Some Initial Observations," Journal of Economic Behavior and Organization, 25 (3), 351-73.

Heide, Jan B. (1994), "Interorganizational Governance in Marketing Channels," Journal of Marketing, 58 (January), 71-85.

Holt, Charles A., Jr. (1980), "Competitive Bidding for Contracts Under Alternative Auction Procedures," Journal of Political Economy, 88 (3), 433-45.

Jap, Sandy D. (1999), "'Pie-Expansion' Efforts: Collaboration Processes in Buyer-Supplier Relationships," Journal of Marketing Research, 36 (November). 461-75.

----- (2002). "Online, Reverse Auctions: Issues, Themes, and Prospects for the Future," Marketing Science Institute-Journal of the Academy of Marketing Science, 30 (Special Issue), 506-525.

Jones, Del (2000), "E-Purchasing Saves Businesses Billions," USA Today, (February 7), 1B.

Jöreskog, Karl G. and Dag Sörbom (1993), LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language. Hillsdale, NJ: Lawrence Erlbaum Associates.

Kagel, John H. (1995), "Auctions: A Survey of Experimental Research," in The Handbook of Experimental Economics, J.H. Kagel and A.E. Roth, eds. Princeton, NJ: Princeton University Press.

----- and Dan Levin (2001), "Behavior in Multi-Unit Demand Auctions: Experiments with Uniform Price and Dynamic Vickrey Auctions," Econometrica, 69 (2), 413-54.

Kisiel, Ralph (2002), "Supplier Group Seeks Conduct Code for Auctions," Automotive News, 67 (5974), 16F.

Lusch, Robert F. and James R. Brown (1996), "Interdependency, Contracting. and Relational Behavior in Marketing Channels," Journal of Marketing, 60 (October), 19-38.

Masten, Scott E. (1988), "Equity, Opportunism, and the Design of Contractual Relations," Journal of institutional and Theoretical Economics, 144 (1), 180-95.

McAfee, R. Preston and John McMillan (1987), "Auctions and Bidding," Journal of Economic Literature, 25 (2). 699-738.

Milgrom, Paul (1989), "Auctions and Bidding: A Primer," Journal of Economic Perspectives, 3 (3), 3-22.

----- and Robert J. Weber (1982), "A Theory of Auctions and Competitive Bidding," Econometrica, 50 (5), 1089-1122.

Noordeweir, Thomas G., George John, and John R. Nevin (1990), "Performance Outcomes of Purchasing Arrangement in Industrial Buyer-Vendor Relationships," Journal of Marketing, 54 (October), 80-93.

Ping, R. and F. Robert Dwyer (1992), "A Preliminary Model of Relationship Termination in Marketing Channels," in Advances in Distribution Channel Research, Vol. 1, G.L. Frazier, ed. Greenwich, CT: JAI Press, 215-33.

Purchasing (2001), "What Top Supply Execs Say About Auctions," (June 21), 52.

Rousseau, Denise M., Sim B. Sitkin, Ronald S. Burt, and Cohn Camerer (1998), "Not So Different After All: A Cross-Discipline View of Trust," Academy of Management Review, 23 (July), 393-404.

Rusbult, Caryl H. and Paul A.M. Van Lange (1996), "Interdependence Processes," in Social Psychology: Handbook of Basic Principles, E.T. Higgins and A.W. Kruglanski, eds. Boston: Guilford Press, 564-96.

Scott-Lewis, David (2001), "Strategic Sourcing + Reverse Auctioning: A Dynamic Duo--Part 1," Research Report No. 1109, Meta Group, (September 7).

Swinkels, Jeroen M. (2001), "Efficiency of Large Private Value Auctions," Econometrica, 69 (1), 37-68.

Tully, S. (2000), "The B2B Tool That Really Is Changing the World," Fortune, 141 (6), 132-45.

Wathne, Kenneth H. and Jan B. Heide (2000), "Opportunism in Interfirm Relationships: Forms, Outcomes, and Solutions," Journal of Marketing, 64 (October), 36-51.

Williamson, Oliver E. (1985), The Economic Institutions of Capitalism, New York: The Free Press.

----- (1993). "Calculativeness, Trust, and Economic Organization," Journal of Law and Economics, 36 (April), 453-86.

----- (1996), The Mechanisms of Governance. New York: Oxford University Press.

The Supplier's Suspicions of Opportunism (α = .79)

(mean = 2.94; s.d. = 1.40; minimum = 1, maximum = 7)

How likely is it that [the buyer firm] would do the following? (1 = "very unlikely"; 7 = "very likely")

Make false accusations.

Provide false information.

Be unwilling to accept responsibility.

Expect my firm to pay for more than their fair share of the costs to correct a problem.

The Supplier's Willingness to Make Idiosyncratic Investments (α = .87)

(mean = 4.50; s.d. = 1.68; minimum = 1, maximum = 7)

In working with [the buyer firm], my firm may have opportunity to make investments in time, energy, and/or money specifically to accommodate it. These investments would be lost if my firm switched to another customer.

Just for [the buyer firm], we would be willing to provide dedicated... (1 = "strongly disagree"; 7 = "strongly agree")

Training for buyers.

Production procedures.

Capital equipment and tools.

Plant capacity.

The correlation between this construct and the supplier's suspicions of opportunism is .12 (p < .17).

Willingness to Collaborate (α = .81)

(mean = 5.08; s.d. = 1.48; minimum 1, maximum = 7)

How willing is your firm to do the following for the buyer? (1 = "very unwilling"; 7 = "very willing")

Participate in product design efforts.

Work together to exploit unique opportunities.

Work on joint projects tailored to their needs.

Look for synergistic ways of doing business.

Satisfaction with the Relationship (α = .87)

(mean = 4.40; s.d. = 1.46; minimum = I, maximum = 7)

(1 = "strongly disagree"; 7 = "strongly agree")

Our relationship with [the buyer] has more than fulfilled our expectations.

We are satisfied with the outcomes of our relationship.

Our relationship with [the buyer] has been a successful one.

The Supplier's Perceived Dependence on the Buyer (α = .90)

(mean = 3.34; s.d. = 1.73; minimum = 1, maximum = 7)

(1 = "strongly disagree"; 7 = "strongly agree")

If our relationship were discontinued with [the buyer], we would have difficulty making up sales volume.

It would be difficult for us to replace [the buyer].

We are quite dependent on (the buyer].

~~~~~~~~

By Sandy D. Jap

Sandy D. Jap is Associate Professor of Marketing, Goizueta Business School, Emory University. This research was supported by grants from the Center for EBusiness at MIT, the Leaders for Manufacturing Program, and the MIT-Ford Alliance. The author thanks the buying organization, sourcing managers, suppliers, and auctioneers for their cooperation throughout the data collection process and John Lynch, Prasad Naik, Nader Tavassoli, Joe Urbany, and the anonymous JM reviewers for helpful comments.

An Investigation into the Antecedents of Organizational Participation in Business-to-Business Electronic Markets 

Business-to-business electronic markets have a profound influence on the manner in which organizational buyers and sellers interact. As a result, it is important to develop an understanding of the behaviors of firms that participate in these markets. The authors develop a typology for the nature of organizational participation to explain the behaviors of user firms in business-to-business electronic markets. The proposed model hypothesizes that the nature of participation depends on organizational motivation and ability. The authors conceptualize motivational factors in terms of efficiency and legitimacy motivations and theorize that ability results from the influence of organizational learning and information technology capabilities. They test the model using organizational-level survey data from jewelry traders that conduct business in an electronic market. The results indicate that both motivation and ability are important in determining the nature of participation; however, the level of influence of motivation and ability varies with the nature of participation.

Electronic commerce is a growing reality and provides viable alternatives to traditional forms of commerce. The major impact of electronic commerce is expected to occur in the business-to-business sector, which is estimated to be approximately six times larger than the business-to-consumer electronic commerce sector, and to reach $1.3 trillion by 2003 (BusinessWeek 2000). Moreover, many large firms recently have announced their ventures into the realm of business-to-business electronic commerce using Internet protocols (communication standards). For example, the recent alliance between the big three auto manufacturers (General Motors, Daimler Chrysler, and Ford) to establish Covisint (an electronic market for automotive parts suppliers), the launch of RetailLink by Wal-Mart, and the Transfer Process Network by General Electric all fall in this category. In addition, in fragmented industries, a host of third-party-operated electronic markets are emerging; more than 750 were in existence at the beginning of the year 2000 (The Economist 2000). Examples include e-STEEL for the steel industry, IMX for home mortgage, and Paper Exchange for the paper industry. The economic significance of electronic markets makes it imperative to study them. Our research takes important steps in this direction by Otudying participant organization behavior in a third-party electronic market.

Despite the interest in third-party electronic markets of practitioners (BusinessWeek 1998) and academicians (Kaplan and Sawhney 2000), concerted efforts to understand these markets have been lacking. Previous research emphasizes the effect of new technologies on organizational processes (Glazer 1991; Heide and Weiss 1995) but does not discuss the influences of electronic markets, and the extant case-based research focusing on electronic markets limits itself to the characteristics of the market maker, the firm that manages and administers the market (e.g., Bakos and Brynjolfsson 1993; Hess and Kemerer 1994). The success of electronic markets, however, depends not only on the characteristics of the market maker but also on the value the market provides for the participant organizations. These participant organizations derive value by isteracting with other participant organizations in the market. Market makers build such positive network externalities by making it attractive for participant organizations to (1) initially adopt the market, (2) subsequently elect to stay on, and (3) be satisfied with the market.

Building positive network externalities, however, depends on the nature of organizational participation, that is, whether the adopting organizations actively participate in the market or are mere passive observers. If a firm is a passive observer, for example, its presence in the electronic market is of no value to other firms in the market because it is unlikely to engage in business transactions with these other firms. The primary purpose of this research is to investigate the factors that influence the nature of organizational participation in electronic markets, which is important because (1) electronic markets are becoming viable alternatives to traditional markets and hierarchies (Bakos 1998); (2) the commercial potential of electronic commerce is immense, and across the globe business-to-business electronic markets are the fastest growing electronic commerce phenomenon (BusinessWeek 1999); and (3) little, if anything, is known about the factors that influence the nature of organizational participation in electronic markets (Alba et al. 1997). Developing an understanding of the nature of organizational participation also would enable market makers to make their markets more attractive to participant organizations.

We rely on the motivation-ability framework (Merton 1957) to develop our conceptual model and make four important contributions. First, we develop and find empirical support for a typology for the nature of organizational participation. We believe that the nature of firm participation is pivotal in developing an understanding of organizational behaviors in electronic markets. Second, we demarcate the two primary organizational motivations for adopting electronic markets; in addition to the traditional efficiency motive, we survey the literature in institutional theory (DiMaggio and Powell 1983) to develop the legitimacy motive. As well as providing realism for the rational models of organizations, the legitimacy motive, together with the efficiency motive, renders a comprehensive conceptualization of organizational motivations. Third, we move beyond case-based research that typically has been used to study electronic markets and use organizational-level survey data to examine our research hypotheses. Fourth, our study tests the motivation-ability theory in the new context of electronic markets. Testing a theory in diverse contexts helps gauge the strengths of the theory and aids in empirical generalizations (Bass 1995).

Time and again, technology has had major influences on marketing practice and scholarship. Recent examples include the influence of integrated information and communication technology (Buzzell 1985). These technologies have revolutionized retailing by the use of scanner data, which also has resulted in improved marketing research models (Blattberg, Glazer, and Little 1994; Curry 1993). In terms of facilitating buyer-seller interactions, electronic data interchange (EDI) helps firms build close relationships and facilitate logistics (Stern and Kaufmann 1985). The EDI systems were generally proprietary to the initiating firm (Frazier 1983) and required significant investments on the part of both firms in a relationship (O'Callaghan, Kaufmann, and Konsynski 1992). Nonproprietary EDI systems were often industry created and facilitated logistic and operational management of interfirm relationships (McGee and Konsynski 1989). In this string of technological advances, the most recent phenomenon is the emergence of the Internet. Internet technology has immense implications for various branches of marketing, including consumer (Hoffman and Novak 1996; Peterson, Balasubramanian, and Bronnenberg 1997), business (Kaplan and Sawhney 2000; Klein and Quelch 1997), and international (Mehta, Grewal, and Sivadas 1996; Quelch and Klein 1996) marketing.

Internet-based business-to-business electronic markets represent an interorganizational information system that facilitates electronic interactions among multiple buyers and sellers (Bakos 1991; Choudhury, Hartzel, and Konsynski 1998). At a basic level, electronic markets can be viewed as information technology (IT)-facilitated markets (Bakos 1998). In electronic markets, buyers and sellers come together in a marketspace and exchange information related to price, product specifications, and terms of the trade, and a dynamic price-making mechanism (such as the bid-and-ask system) facilitates transactions between the firms (Kaplan and Sawhney 2000). Depending on the configurations of buyers and sellers and the price-making mechanism used, such electronic markets have been referred to as catalog aggregators, auctions, reverse auctions, or exchanges. Unlike EDI, which helps maintain existing interfirm relationships, electronic markets help buyers search for sellers and sellers search for buyers to engage in transactions. Thus, although EDI helps firms achieve strategic objectives by facilitating operational management of buyer-seller relationships, electronic markets are more strategic and assist firms to interact with other firms in a market setting (as opposed to a relational setting).

Usually an electronic market is sponsored or maintained by a market maker. The primary function of market makers is to gather buyers and sellers in a marketspace (Klein and Quelch 1997). Market makers also may perform other supplementary functions, such as providing credit, warranties, and logistics, and sometimes even may help in negotiations among potential business partners. The business models for market makers tend to vary with the functions they perform in addition to providing the marketspace (Kaplan and Sawhney 2000).

Depending on the position of the market maker in the value chain, an electronic market may be either hierarchical (biased) or market-driven (third-party) (Malone, Yates, and Benjamin 1994). In a hierarchical electronic market, the market maker is also a buyer or a seller. An example of such an environment is a seller-sponsored market such as Sabre (initially sponsored by American Airlines). Hierarchical markets are inherently biased toward the sponsor (i.e., the market maker), and as a result the market maker has advantages over competitors that conduct business in that market. Copeland and McKenney's (1988) detailed study of airline reservation systems (Sabre) points to these disadvantages of biased electronic markets, which eventually limit the scope and the survival likelihood of such markets.

A third party (neither a buyer nor a seller) sponsors an unbiased market-driven electronic market; thus, the market maker does not carry out transactions in the market. Examples include PaperExchange for paper and related products and SportsNet, which links more than 3500 dealers of sports cards. It is these third-party markets that have experienced the most success and are forecasted to dominate fragmented industries (BusinessWeek 1999; Krantz 1999). In our research, we study buyers, sellers, retailers, pawnbrokers, appraisers, and other intermediaries that trade in jewelry and related products in a third-party, unbiased, arket-driven electronic market, Polygon, as we discuss in greater detail subsequently (henceforth, we use "organization" and "firm" to refer to a user firm that participates in an electronic market).

In this section, we provide a rationale for a firm's need to participate in electronic markets and then present background on the nature of organizational participation, the motivation and ability variables, and environmental dynamism, the dominant characteristic of electronic markets. Strategic considerations motivate organizations to build capabilities and preempt competition and thereby to serve customers better (Day and Wensley 1988; Slater and Narver 1995). Considering the prominent role of technology in modern society (Blattberg, Glazer, and Little 1994; Buzzell 1985), it comes as no surprise that organizations view technology as a means of building sustainable competitive advantage (Day and Glazer 1994; Glazer and Weiss 1993). We contend that strategic considerations, such as providing better customer service (Parasuraman and Grewal 2000) and competitive hedging and/or preemption (Dickson 1992), gain significance in the organizational decision to participate in electronic markets.

We theorize that the nature of organizational participation in an electronic market depends on a firm's motivation and ability (Figure 1). Literature in diverse fields such as consumer behavior (MacInnis, Moorman, and Jaworski 1991), organizational behavior (O'Reilly and Chatman 1994), and marketing strategy (Boulding and Staelin 1995) emphasizes the criticality of both motivation and ability. However, we broaden the theorizing on motivations to encompass (1) the traditional emphasis on efficiency (Rindfleisch and Heide 1997) and (2) the institutional motivations of attaining legitimacy, which we refer to as the legitimacy motive (DiMaggio and Powell 1983). Furthermore, we rely on the organizational learning literature (Sinkula 1994) to conceptualize a firm's ability as its (1) learning about the electronic market and (2) IT capabilities. We also study the influence of environmental dynamism that is inherent in electronic markets (Lee and Clark 1997).

Nature of Organizational Participation

Organizations participate in electronic markets in several ways. At one extreme, organizations spend concerted effort and resources to streamline their electronic market operations. At the other extreme, they adopt electronic markets on an experimental basis (Dickson 1992). We develop a istinct-states conceptualization for the nature of organizational participation to illustrate diverse organizational behaviors in electronic markets. In addition to the literature on innovation assimilation (Meyer and Goes 1988) and electronic markets (Choudhury, Hartzel, and Konsynski 1998), we rely on exploratory interviews to develop our typology for the nature of organizational participation. Several times during the research process, we interviewed the chief executive officer (CEO) and marketing manager of the market maker. The CEO noted that customers were "a mixed lot with many different activity levels." Managers of participant firms also suggested that participants tend to have varying levels of activities in the electronic market.

To capture the varying levels of firm activity in electronic markets, we conceptualize the nature of firm participation in terms of the exploration state, the expert state, and the passive state. In the exploration state, firms do not know the requirements to conduct operations in an electronic market effectively but are expending substantial efforts to learn the particulars of doing business in the marketplace. Firms in the exploration state are "testing the waters" to understand the new medium better and in the process are using cognitive, physical, and financial resources. These organizations are trying to sense which business practices they need to reengineer, how they can reengineer, and whether it will be in their best interest to reengineer.

In the expert state, firms believe they have been successful in reengineering their business processes to function effectively in the electronic market. Similar to experts, they possess the know-how to perform market-related tasks successfully (Alba and Hutchinson 1987). Expert firms have substantive knowledge about their electronic market and procedural knowledge pertaining to the way of doing business in the market. These firms also understand the cause-and-effect relationships for their activities in the electronic market (i.e., axiomatic knowledge) and are thereby in a position to make high-quality decisions regarding their market operations (Sinkula 1994; Slater and Narver 1995). As the electronic market evolves, the expert firms regularly update their knowledge base about the market to remain current and thereby create episodic knowledge (Sinkula 1994). In a manner, expert firms have made an implicit or explicit pledge to continue conducting business over the electronic medium.

In the passive state, organizations carry out virtually no business in the electronic market but continue to maintain a presence. Our exploratory research suggests that the passive state is (1) propagated by firms entering electronic markets on an experimental basis, perhaps to supplement their traditional markets at some time in the future; (2) due to competitive hedging wherein a firm does not believe that electronic markets are viable but considers them to be a future opportunity or threat and therefore wants to observe and learn; (3) perpetuated by low entry barriers, in that entering an electronic market involves buying a computer to access the Internet; and (4) reinforced because maintaining a presence in business-to-business electronic markets is not expensive, requiring a firm simply to pay its monthly subscription fee. Organizations in this state are unwilling to expend the cognitive, physical, or financial resources that are needed to develop human capital and reengineer business practices to conduct business over the electronic medium actively.

Motivation

The literature on organizational founding suggests that the motives, processes, and structures that firms stress at the time of their inception have a long-lasting and perpetual influence on their behaviors (Baum and Oliver 1992; Schulz 1998). Accordingly, we expect the motives a firm emphasizes when entering an electronic market to influence the firm's operations in the market for a substantial period. We suggest that organizational motivations for entering electronic markets include an economic expectation of enhancing efficiency and a normative objective of attaining legitimacy.<SUP>4</SUP> We rely on transaction cost economics to develop the efficiency motive (Rindfleisch and Heide 1997), and the literature on institutional theory provides the bases for the legitimacy motive (Meyer and Rowan 1977).

Efficiency motive. An economist would argue that electronic markets improve transaction effectiveness and efficiency (Rindfleisch and Heide 1997). Improving efficiency is likely to be driven by the organizational strategic consideration of serving customers better (Day and Nedungadi 1994; Day and Wensley 1988). As Dickson (1992) observes, to serve their customers effectively and efficiently, firms often are motivated to experiment with new ways and innovations. Malone, Yates, and Benjamin (1994) suggest that electronic commerce leads to greater use of markets, rather than hierarchies, because these markets have relatively lower transaction costs. Research in the management of information systems and the economics of electronic markets supports this assertion (Bakos and Brynjolfsson 1993; Gurbaxani and Whang 1991; Hess and Kemerer 1994).

Reducing the cost of doing business is consistent with an organization's attempt to improve efficiency. Recent theoretical developments in transaction cost economics (TCE), which emphasize minimizing transaction costs, explicitly recognize this efficiency orientation and conceptualize TCE as a constrained-efficiency framework (Roberts and Greenwood 1997; Williamson 1992). This perspective regards organizations as efficiency seeking (Nelson and Winter 1982) under cognitive constraints (e.g., bounded rationality; Williamson 1987) and/or institutional restraints (Scott 1987).

Efficiency considerations are usually internal to an organization, but often attempts to enhance efficiency are externally oriented (Oliver 1990). The efficiency motive of reducing the costs of transacting with vendors, for example, might prompt an organization to use automated systems. In addition, the efficiency motive can be used to gauge the emphasis an organization puts on reducing costs and enhancing productivity (internal consideration) when entering an electronic market (external orientation). Over time, this emphasis on efficiency should become embedded in the organizational culture and influence the formal and informal functioning of firms (Deshpande and Webster 1989). This goal-directed orientation also should increase organizational commitment toward IT and organizational effectiveness in electronic markets (Ginzberg 1981; Newman and Sabherwal 1996).

Legitimacy motive. Institutional theory suggests that organizations must justify their actions and perform in accordance with existing societal norms and institutional expectations (DiMaggio and Powell 1983; Scott 1987). Adhering to societal norms enhances an organization's legitimacy and increases the likelihood of organizational survival. One way for organizations to attain legitimacy is to carry out activities that are deemed suitable by institutional constituents, including the government, consumer bodies, trade associations, and the public.

Electronic markets by their very nature are technologically intensive, and organizations dealing in these environments are likely to be perceived as having technological savvy. Being perceived as technologically knowledgeable is a definite advantage (Glazer and Weiss 1993). Thus, a possible organizational motive to enter electronic markets is to portray an image of technological sophistication. In other words, organizational stakeholders view technologically sophisticated firms more favorably in comparison with technologically naive firms; therefore, the organizational legitimacy of technologically sophisticated firms is higher because they provide a better fit with the modern-day organizational profile.

Organizations also mimic behaviors of a successful benchmarked group (Deephouse 1996; DiMaggio and Powell 1983). Institutional theorists argue that imitation is an uncertainty reduction mechanism in the sense that when a firm successfully adopts a structural change, other organizations mimic the change while attributing their success to the nature of the structural transformation (Haunschild and Miner 1997; Haverman 1993). Firms usually mimic the structures and the processes of other firms that are perceived to be legitimate (Haunschild 1993; Suchman 1995). Diverse research streams, including those on strategic alliances (Pangarkar and Klein 1998) and outsourcing (Lacity and Hirschheim 1993), have studied this mimicking behavior under the rubric of the bandwagon effect. The bandwagon effect suggests that sometimes organizations engage in activities simply because other firms do and provides a possible reason for the prominence of strategic alliances in some industries and their absence from others (Osborn and Hagedoorn 1997). In this study, organizational attempts to appear technologically sophisticated and mimic the behavior of successful organizations are classified as the legitimacy motives.

Ability

Knowledge acquisition and utilization processes help firms build capabilities and sustain strategic competence (Fiol and Lyles 1985). Knowledge developmental processes also provide organizations with resources, assets, and skills to compete effectively (Sinkula 1994; Slater and Narver 1995). Our research studies (1) organizational learning, a major source for building a knowledge base, and (2) a firm's IT capabilities, an avenue for extracting rents from the knowledge base (March 1991). By virtue of the newness and novelty associated with electronic markets, organizations have either not developed business models for this medium or tested existing business models in the medium. Learning therefore becomes pivotal for developing and adopting business models. Organizational capabilities reflect the extent of knowledge utilization in a firm and are critical for building sustainable competitive advantage (Day 1994). In electronic markets, a dominant organizational resource is a firm's IT capabilities (Guha et al. 1997).

Learning. Huber (1991, p. 90) describes knowledge acquisition as "the process by which knowledge is obtained," and this learning by an organization is a function of its age and effort (Garvin 1993). In organizations, experience is a prime source of learning and captures the incessant trial-and-error process by which organizations acquire information (Sinkula 1994). This acquisition of information leads to richer and proprietary knowledge bases. Eventually, the distribution, interpretation, and utilization of knowledge bases result in sustainable competitive advantages.

We examine two important learning constructs. The first is age-based or experiential learning, which we define as the learning an organization obtains from the extent of its experience and operationalize as the age of a firm's electronic market operations. Experiential learning has been shown to predict organizational survival, even after size and resource-enhancing linkages are controlled for (Brittain 1989), and the literature on operations management documents the positive effect of age on performance (Mody 1989). Although age is an important indicator of learning, it does not capture the effort spent in learning. To overcome this weakness, our second learning construct taps directly into the effort a firm devotes to developing skills to manage an electronic market. We refer to this construct as effort-based learning and conceptualize it as the effort, in terms of organizational resources and human capital, used to develop knowledge about an electronic market (Simon 1991).

IT capabilities. Firm capability hinges on the efficient development and use of resources (Sinkula 1994). Information use has been the focus of marketing research since the American Marketing Association/Marketing Science Institute-sponsored workshop on knowledge development (Myers, Massy, and Greyser 1980). This research stream indicates that knowledge development enhances the value of firms' resources and organizational capabilities (Moorman, Zaltman, and Deshpande 1992). In electronic markets, a participant's IT capability is an important organizational resource, should play a vital role in building sustainable competitive advantages, and should increase the firm's capacity to manage electronic markets (Auer and Reponen 1997; Mata, Fuerst, and Barney 1995).

Organization information processing research suggests that it is not organizational capabilities per se but the fit between organizational information processing needs and capacity that is critical (Galbraith 1973; Tushman and Nadler 1978). Information serves as a means of reducing uncertainty and thereby attaining the desired objective. In turn, information requirements depend on uncertainty, and by extension the desired information processing capacity depends on the information processing needs.

Environmental Dynamism

Environmental dynamism is a dominant characteristic of electronic markets (Klein and Quelch 1997). Rapidly increasing subscription bases, along with ever-changing technologies, characterize electronic markets (Lee and Clark 1997). Environmental dynamism captures these changes in demand and technology (Weiss and Heide 1993), induces uncertainty (Achrol and Stern 1988), and influences organizational structures and processes (Achrol 1991).

Because we have conceptualized the nature of organizational participation in terms of three distinct states (exploration, expert, and passive), we need to use one of the states as a base state and compare it with the other states. That is, because a firm can be in only one state, the three states compete against one another in some fashion. In formulating our hypotheses, we use the exploration state as the base state because initially almost every firm should be in this state.

Efficiency Motive and IT Capabilities

According to the motivation-ability framework, both firm motivation (efficiency motive) and ability (IT capabilities) should be present to affect organization participation (Merton 1957). Research shows that information systems projects often fail when either motivation or ability is lacking (see Ewusi-Mensah and Przasnyski 1991; Reich and Benbasat 1990). We therefore expect both motivation and ability to influence the nature of organizational participation. Emphasis on the efficiency motive and IT capabilities should enhance the likelihood of a firm being in the expert state and lower the likelihood of the firm being in the passive state.

H1: The greater the emphasis on the efficiency motive and IT capabilities, (a) the higher is the likelihood of a firm being in the expert state and (b) the lower is the likelihood of a firm being in the passive state.

Organization information processing research suggests that the influence of IT capabilities is likely to be nonlinear (Daft and Lengel 1986). As Day (1994) observes, the market dictates the extent and nature of capabilities that an organization develops, and the information processing needs determine the extent of information processing capacity required (Day and Glazer 1994). As organizational information processing capacity increases, the need to develop more of this capability declines (March 1991). In other words, to a specific level, IT capabilities should help a firm reach the expert state and reduce the chances of the firm being in the passive state, but eventually the influence of IT capabilities should level off. We therefore hypothesize that as the level of IT capabilities increases, the likelihood of a firm being in the expert state increases, but at a declining rate. We also propose that as the level of IT capabilities increases, the likelihood of a firm being in the passive state decreases, but at a declining rate.

H2: As the level of IT capabilities increases, IT capabilities (a) positively affect the likelihood of a firm being in the expert state, but at a declining rate, and (b) negatively affect the likelihood of a firm being in the passive state, but at a declining rate.

Legitimacy Motive

Legitimacy motives represent organizational attempts to adopt electronic markets either to portray a specific image to stakeholders or to mimic benchmarked organizations. Legitimacy is construed as "a generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially construed system of norms, values, beliefs, and definitions" (Suchman 1995, p. 574). It seems that if enhancing organizational legitimacy is the motive for joining an electronic market, an organization achieves its objective just by entering the market. In other words, by virtue of a firm's entry into an electronic market, it is in a position to assert to its stakeholders that it is technologically advanced and ready for the challenges of the information age. Organizations that embrace electronic markets to mimic a successful benchmark firm believe that the benchmarked organization succeeded primarily because of its participation in electronic markets. In summary, some organizations ceremoniously adopt electronic markets as a mere pretense to attain legitimacy (DiMaggio and Powell 1983; Scott 1987).

For firms that emphasize the legitimacy motive, attempts to acquire knowledge or build capabilities are likely to be minimal. Reaching the expert stature without knowledge development is unlikely (Alba and Hutchinson 1987). In other words, simply entering electronic markets may not be sufficient to attain the expert state, which requires a firm to learn new routines, rules, and strategies. Attaining legitimacy in the eyes of stakeholders and adopting electronic markets on an experimental basis are likely to motivate a firm to maintain presence in the electronic market, even if the presence does not yield direct economic gains. Therefore, emphasis on legitimacy motives should lower the likelihood of a firm being in the expert state and increase the likelihood of the firm being in the passive state.

H3: The more an organization subscribes to legitimacy motives for entering an electronic market, (a) the lower is the likelihood of a firm being in the expert state and (b) the higher is the likelihood of a firm being in the passive state.

Learning

Research on organizational learning suggests that age-based learning helps capability development (Sinkula 1994). However, such learning has been shown to have limitations related to unintentional and unsystematic learning. As Levinthal and March (1993, p. 96) observe, "Experience is a poor teacher, being typically quite meager relative to the complex and changing nature of the world in which learning is taking place." Frequently, employees are the sensors that gather information and draw inferences. The cognitive constraints of rationality that inhibit human activity are also likely to impede age-based learning (Dickson 1992). Experience also tends to be ambiguous, with multiple and conflicting interpretations, making it difficult to decipher the "complex worlds" (Levinthal and March 1993, p. 97). Research on experience curves attests to this supposition and demonstrates that learning increases with age, but at a decreasing rate (Bass 1995). We expect age-based learning to increase the likelihood of a firm being in the expert state, but the rate of increase should lessen with age. Firms that are not effective learners, however, should sooner or later move to the passive state, and the probability that these firms will move into the passive state should vary positively with the age of their electronic market operations.

H4: As the level of age-based learning increases, age-based learning (a) positively affects the likelihood of a firm being in the expert state, but at a declining rate, and (b) negatively affects the likelihood of a firm being in the passive state at an increasing rate.

In addition to age-based learning, the effort that an organization spends in developing its knowledge bases is an important aspect of organizational learning. However, effort does not imply automatic success; organizations often spend considerable efforts in new ventures and fail (Golder and Tellis 1993). In such cases, firms fail to develop the skills and knowledge bases needed to manage the new ventures (Tushman and Anderson 1997). If effort-based learning is successful, it helps firms build their skill levels and move on to the expert state. In contrast, if effort fails to develop knowledge, the commitment to a failing course of action is likely to be minimal, and the level of effort should decline. Therefore, the probability of the firm moving to the passive state should increase.

H5: As the emphasis on effort-based learning increases, (a) the likelihood of a firm being in the expert state increases and (b) the likelihood of a firm being in the passive state increases.

Environmental Dynamism

Dynamic environments are characterized by high variability in demand and unpredictability regarding the actions of competitors (Achrol and Stern 1988). Research suggests that environmental dynamism makes it difficult for organizations to assimilate and anticipate environmental conditions and has an adverse influence on performance (March 1991). These challenges make it hard for firms to identify and develop the skills needed to succeed. As a result, learning is likely to be a longer and a more deliberate process. We therefore expect environmental dynamism to reduce a firm's likelihood of being in the expert state. Environmental dynamism is also likely to frustrate firms and therefore increase the likelihood that the firms give up in dynamic markets, thereby enhancing the probability of the firms being in the passive state.

H6: The higher the environmental dynamism in an electronic market, (a) the lower is the likelihood of a firm being in the expert state and (b) the higher is the likelihood of a firm being in the passive state.

Firm-specific factors typically moderate the influence of environmental variables (Dess and Beard 1984). In electronic markets, IT capability is one such firm-specific variable that should help organizations manage environmental dynamism; it is an organizational resource and reflects a firm's ability to exploit accumulated information. Organization information processing theory (Daft and Lengel 1986) also suggests that IT capabilities help manage environmental dynamism and therefore should moderate its effect.

H7: Information technology capabilities will moderate (a) the negative influence of environmental dynamism on the likelihood of a firm being in the expert state and (b) the positive influence of environmental dynamism on the probability of a firm being in the passive state.

Research Context

In our research, we study the nature of participation of firms that adopt Polygon, a market-driven electronic market for jewelry and related products (www.polygon.net). Polygon Network Inc. was founded in 1983, primarily to provide an electronic market for jewelry pawnbrokers, appraisers, manufacturers, wholesalers, retailers, buyers, and sellers. It is a subscription-based service for which members pay a monthly access fee, which enables them to buy and/or sell such jewelry items as cut diamonds, watches, and rings, as well as acquire other benefits such as gemological and related technical information. Polygon migrated in the late 1980s to an exchange format and facilitated interfirm transactions by having user firms call the Colorado server to retrieve their e-mail and information from the buyer-seller bulletin board. In 1995, Polygon acquired the domain name polygon.net and began its migration to Web-based operations. Although the transition to the Web has eased access for user firms, the nature of the information exchange has remained the same. Polygon represents an "open bazaar" in which buyers and sellers meet in an electronic environment and that makes it efficient for firms to exchange information related to price, product specifications, and terms of trade. Although Polygon does not enable participating firms to make payments electronically, it provides ratings for all participating firms based on their payment history.

Independent Measures

Following standard psychometric procedures (Churchill 1979), we created a pool of items for each of the constructs (all items were measured on a five-point semantic differential scale, where 1 = "disagree" and 5 = "agree"), with the exception of age-based learning. We extensively field tested the items by means of personal interviews with Polygon subscribers.

The efficiency motivation construct gauged the degree to which a firm stressed cost reduction and output enhancement when joining Polygon. We used four items that measured organizational emphasis on (1) increasing efficiency, (2) reducing costs of doing business, (3) streamlining operations, and (4) reducing the costs of transacting with exchange partners. The first two items for the legitimacy motivation construct measured the extent to which the firm mimicked competitors and other jewelers, respectively; the third item emphasized legitimacy attainment as a motive; and the final item gauged the extent to which the organization wanted to portray the image of being high tech. We measured a subscriber's IT capabilities by adapting six items from King and Teo's (1996) study of facilitators and inhibitors for the strategic use of IT. To assess effort-based learning, we used a two-item measure. The first item gauged the time and effort expended by employees to learn about Polygon. The second item measured the overall organizational effort. Finally, we adopted Klein, Frazier, and Roth's (1990) environmental dynamism scale.

We used the age of an organization's Polygon operations to assess age-based learning (Brittain 1989). Another instrument for age-based learning could be the number of years in business, not the tenure of involvement with an electronic market. We believe that the tenure of involvement is a more suitable instrument for at least two reasons. First, electronic markets are quite different from traditional forms of commerce, and therefore learning about them is likely to be critical. Second, years in business induces other effects, such as creating inertia (Chandrashekaran et al. 1999), thereby making it difficult to decipher the effect of learning about the electronic market. Bricks-and-mortar retailers, for example, were slower to establish their electronic retail outlets than were new start-ups. However, years in business is an important variable, and further research should examine the dynamics propagated by it.

Dependent Measure: The Nature of Organizational Participation

We used a polychotomous dependent variable to measure the nature of organizational participation. While developing this measure, we assessed the finest possible classification for the nature of organizational participation. This measurement serves two purposes: (1) It helps us rule out the possibility of a fourth state, and (2) it provides flexibility because it is easier to aggregate data than to disaggregate them. The dependent measure asked respondents to choose one of six categories that best described their present Polygon operations. Table 1 details the specific statements for each category. We used two statements to measure the exploration state, one to categorizes the expert state, and three to measure the passive state.

Pretest

To obtain a preliminary assessment of the internal validity of our measurement instrument, we mailed 300 questionnaires to Polygon subscribers and received 34 complete and usable responses. We used item-to-total correlation to assess the validity of the measurement instruments. These correlation coefficients were greater than .78 for all items, and therefore we retained all items. In the final study, we mailed the questionnaire to the remaining 1846 Polygon members and followed the same procedure as in the pretest.

Sample and Nonresponse Bias

We mailed the questionnaire to all 2146 Polygon subscribers in our sample frame and received 306 responses. The survey consisted of the scale items under investigation and a cover letter stating the purpose of the study and that Polygon Network Inc. endorsed the study. The cover letter noted that the purpose of the study was to research organizations at the frontier of adopting the Internet and carrying out electronic commerce. Polygon Network Inc. supplied the name of the key respondent for each firm, and typically the key respondent interacted with the Polygon Network Inc. on a regular basis.

We used the c2statistic to compare the sample (both pretest and final study combined) characteristics with those of the population (all Polygon subscribers) (Table 2). The results showed no significant differences between the characteristics of the sample and the population (c2 = .875, degrees of freedom [d.f.] = 3, p > .83). We compared the responses to our independent variables for the early respondents with those for the late respondents and the pretest respondents with those for the respondents in the final study and found no statistical differences (Armstrong and Overton 1977).

Measure Validation

We carried out measure validation in two phases. First, we personally interviewed three Polygon members and the CEO and marketing manager of Polygon Network Inc. to determine what it meant to do business in this electronic market. On the basis of our discussions, we sought to refine our conceptual model and develop a sample list of items. Then we conducted a second round of interviews to verify our conceptual model and refine our measures. Second, we used confirmatory factor analysis (CFA) to establish the convergent validity for the efficiency motive, legitimacy motive, effort-based learning, IT capabilities, and environmental dynamism scales (Table 3).

We retained all items for the efficiency motive, effort-based learning, and dynamism scales. The item deleted from the legitimacy motive scale emphasized mimicking competitors. Two items were deleted from King and Teo's (1996) scale. The four remaining items measured a firm's (1) IT planning capabilities, (2) strength of technical support staff, (3) understanding of benefits from the application of IT, and (4) knowledge about IT. Our CFA results show that all factor loadings were greater than the recommended .4 cut-off and were statistically significant (Nunnally and Bernstein 1994). The c2 statistic was not significant, which implies that the sample and estimated covariance matrix were alike. The goodness-of-fit index, adjusted goodness-of-fit index, nonnormed fit index, and comparative fit index were greater than the recommended .9; the parsimony normed fit index was greater than the recommended .6; and the root mean square error of approximation, as recommended, was less than .08 and not statistically different from .05 (Hair et al. 1995). To assess the validity of our measure for age-based learning, we used its correlation with (1) the number of transactions made per week on the Polygon and (2) the annual dollars transacted through Polygon. As we expected, these correlation coefficients were positive and statistically significant (r = .296, p < .01, n = 171; r = .191, p < .02, n =171, respectively)

To establish the internal consistency of our measurement model, we examined the reliabilities and average variance extracted (Fornell and Larcker 1981). All the reliabilities were greater than the recommended .7 (Nunnally and Bernstein 1994). The average variance extracted for each measure was greater than the recommended .5, with the exception of the scale for legitimacy motivation, which extracted 46% of variance (Bagozzi and Yi 1988). Additional research could refine our measure of the legitimacy motive. We used two methods to assess discriminant validity. First, the 95% confidence bands around the fs did not contain 1 (Anderson and Gerbing 1988). Second, we compared the average variance extracted with the fs (Fornell and Larcker 1981). The average variance extracted was greater than the respective fs for all measures. Table 4 displays the descriptive statistics.

We used a series of t-tests to examine the differences across the three states for (1) the number of transactions per week, (2) the percentage of business transacted through Polygon, and (3) the time since the adoption of Polygon. We expected the number of weekly transactions to be higher for both the exploration and the expert state in comparison with the passive state, which was the pattern of results we obtained. For the exploration versus passive comparison, we obtained b = 2.30, p < .01, and for the expert versus passive comparison, we obtained b = 2.59, p < .01. A comparison of the exploration and expert states gave us b = -.29, p > .78 (although this test was not significant, in absolute terms, the number of weekly transactions was higher for the expert state). Thus, in terms of number of weekly transactions, the exploration and expert states were statistically equal to each other but higher than the passive state.

As with the number of transactions, the percentage of business that firms undertook through Polygon in the passive state (2.58%) was statistically lower than that of firms in either the exploration (18.31%; b = 15.73, p < .01) or the expert (19.95%; b = 17.37, p < .01) state. Again, although firms in the expert state transacted more business over Polygon than did firms in the exploration state, statistically the two values were equal (b = -1.65, p > .70).

Typically, firms enter the exploration state and then move to either the expert or the passive state. In terms of average time since the adoption of Polygon, our results supported this assertion. Specifically, the age of Polygon operations of firms in the exploration state was statistically lower than that of firms in either the expert (b = -9.03, p < .07) or the passive (b = -11.50, p < .05) state. In addition, there was no statistical difference between firms in the expert state and those in the passive state (b = -2.46, p > .77).

Taken together, the results suggest that the three states differ from one another in expected manners. The exploration state differed from the expert state in terms of time since adoption of Polygon, and both of these differed from the passive state in terms of the number of weekly transactions and the percentage of business transacted over Polygon.

Estimation Model

Because we have a discrete dependent variable with three states (i.e., exploration, expert, and passive), we used a multinomial logit (MNL) model to test our hypotheses.

Where ORG-PAR is the polychotomous dependent variable with the exploration state as the base, EFF stands for efficiency motive, LEGIT represents the legitimacy motive, AGE denotes age-based learning, EFFORT represents effort-based learning, IT-CAP depicts IT capabilities, and ENV-DYN designates environmental dynamism. The maximum likelihood estimate for this model yields two sets of estimates, one for the expert state and the other for the passive state.

Overall Model Test

In Table 5, we display the results from our MNL model with the exploration state as the base. The likelihood ratio test for the overall fit of the model indicates that the independent variables explain statistically significant variance in the dependent measure (c<SUP>2</SUP> = 58.06, d.f. = 20, p < .01). In addition to establishing descriptive validity for the structure of the model, it is also important to evaluate the predictive validity of the model. We used estimates from the MNL model to carry out a discriminant exercise and tabulate the proportion of correct predictions (Table 5). The proportion of correct classifications for the model is 54.0%, which is greater than the two benchmarks recommended by Morrison (1969): maximum chance criterion (45.6%) and proportional chance criterion (35.8%)

Because we have interaction terms in our model, we interpret the main effects as contingent on the appropriate interaction term (Jaccard, Turrisi, and Wan 1990). The main effect of efficiency, for example, represents the impact of efficiency when IT capabilities equal zero. Mathematically then, y = b0 + b1 EFF + b9 EFF IT-CAP, and when IT-CAP = 0, y = b0 + b1 EFF. Therefore, in a way, b1 by itself is meaningless, because the influence of efficiency varies with IT capabilities. To aid in interpreting these coefficients, we carried out Wald tests for the significance of the efficiency motive and environmental dynamism at various levels of IT capabilities and for the significance of IT capabilities at various levels of the efficiency motive and environmental dynamism (Table 6).

Antecedents of the Expert State

The first hypothesis suggests that both IT capabilities and an efficiency motive would be needed for a firm to be in the expert state. We find support for this assertion (b9 = .879, p < .01). To provide a better understanding of this interaction effect, we examined the efficiency motive at three levels of IT capabilities. As Table 6 shows, the influence of the efficiency motive on the likelihood of a firm being in the expert state increases from -1.178 (p < .01) at low levels of IT capabilities to .266 (p > .46) at high levels of IT capabilities. H2 posits that, as IT capabilities increase, the positive effect of IT capabilities on the likelihood of a firm being in the expert state will decline. We do not find support for this hypothesis (b7 = -.108, p > .33). Our hypothesis on the legitimacy motive, which proposes that an emphasis on the legitimacy motive will decrease the likelihood of a firm being in the expert state, is marginally supported (H3 : b2 = -.355, p < .10). H4 suggests that age-based learning tends to increase the likelihood of a firm being in the expert state and that this positive effect diminishes as the level of age-based learning increases. Our data support this hypothesis, with significance of both the linear (b3 = .473, p < .01) and square (b4 = -.024, p < .05) terms for age-based learning. As Figure 2 shows, the influence of age-based learning increases at a decreasing rate. H5 posits a positive relationship between effort-based learning and the likelihood of a firm being in the expert state. We find marginal support for this hypothesis (b5 = .389, p < .10). We hypothesized in H6 that environmental dynamism would have a negative effect on the probability of a firm being in the expert state and in H7 that IT capabilities would moderate this negative effect. Our results support both these hypotheses (H6 : b8 = -2.294, p < .05; H7 : b10 = .757, p < .05). As Table 6 shows, the effect of environmental dynamism varies from a low of -.197 (p > .62) at low levels of IT capabilities to a high of 1.012 (p < .05) at high levels of IT capabilities.

Antecedents of the Passive State

H1 posits that emphasis on both IT capabilities and efficiency motives should decrease the likelihood of a firm being in the passive state. Our results do not support this hypothesis (b9 = .294, p > .12). The main effect for the efficiency motive (b1 = -1.858, p < .05) shows that such an emphasis reduces the likelihood of a firm moving to the passive state. H2 proposes that the IT capabilities will negatively influence the probability of a firm being in the passive state and that this negative influence will decrease as the level of IT capabilities increases. Our results support this hypothesis because both the linear (b6 = -3.836, p < .05) and square (b7 = .247, p < .10) terms are significant. H3 suggests that the legitimacy motive will have a negative effect on the probability of a firm being in the passive state. Our data do not support this hypothesis (b2 = -.203, p > .17). We expected the negative effect of age-based learning on the likelihood of a firm being in the passive state to increase with the level of age-based learning. Our results do not support this hypothesis, because neither the linear (b3 = .074, p > .16) nor the square (b4 = .001, p > .31) terms are significant. We hypothesized in H5 that effort-based learning would move a firm away from the exploration state, and we find support for this hypothesis (b5 = .293, p < .10).<SUP>6</SUP> Environmental dynamism was hypothesized to influence the probability of a firm being in the passive state positively, and we expected this positive effect to be moderated by IT capabilities. We do not find support for either the main effect hypothesis (H6 : b8 = -.794, p > .22) or the moderating role of IT capabilities (H7 : b10 = .371, p > .11).

Our findings suggest that participating firms must emphasize the right motives and should have the requisite capabilities to participate actively in electronic markets. An emphasis on the efficiency motive and IT capabilities is critical for firms to attain the expert state and avoid the passive state. There is, however, a critical difference between the results for the expert state and those for the passive state. For the expert state, the interaction term between the efficiency motive and IT capabilities is significant, whereas this interaction term is not significant for the passive state. In other words, the influences of the efficiency motive and IT capability on the likelihood of a firm being in the passive state are independent of each other, whereas in the case of the expert state, the two variables act as catalysts in attenuating each other's influence. The finding that IT capabilities help firms manage the dynamism inherent in electronic markets further strengthens the criticality of this organizational ability. However, our results do not support the hypothesized nonlinear effect of IT capabilities, perhaps because of the context of our study. The jewelry business is not traditionally high tech; consequently, our sample may have firms that have not reached saturation in terms of IT skills. A replication of our study in a high-tech industry or with a multi-industry sample may provide for a more thorough testing of the nonlinear effect of IT capabilities.

In terms of motivation, developing a proper mindset by stressing efficiency motives and de-emphasizing legitimacy motives is critical for firms to attain the expert state. To become experts, firms also must emphasize organizational learning (age- and effort-based). However, the positive influence of age-based learning tends to decline with age. Although effort-based learning is critical for attaining the expert state, it also tends to increase the likelihood of a firm becoming a passive participant. Overall, we find marginal support for our assertion that effort-based learning will move firms out of the exploration state.

Managerial Implications

Firms that intend to participate in electronic markets should be aware that their nature of participation is dependent not only on their IT capabilities but also on their motivation. Entering the market simply to jump on the bandwagon or establish an image of being technologically proficient does not seem to enhance activity levels in the market. Moreover, to become experts, firms should strive to achieve efficiency and work to build their IT capabilities. Allocating time and effort to learn and understand the environment also can yield substantial benefits. Firms that enter electronic markets on an experimental basis or with the desire to mimic others will most likely become passive participants. If a firm joins an electronic market to experiment, it must expend resources to achieve the objectives of its experiment.

Market makers are more likely to succeed if they provide positive network externalities for participating firms. A market maker must understand the reasons for a user firm's participation and thereby be able to provide the right incentives for these firms to adopt the market. Understanding the goals of their participating firms enables market makers to design programs that facilitate goal achievement and enhance member retention. Polygon, for example, could identify active firms and encourage them to develop their IT capabilities, as well as facilitate their learning about the environment. Passive participants might be encouraged to retain their membership for reasons that reinforce their objectives. However, convincing the passive participants to expend resources to become experts may be in the best interest of a market maker in the long run. A market maker's strategic orientation also influences the type of companies that join the market. If a market maker wants to create a market primarily with expert firms, for example, its strategic orientation and tactical actions should identify IT-capable firms whose objectives are to make the market their major sales channel.

Research Implications

Our primary contribution to theory lies in developing and substantiating the existence of a typology for the nature of organizational participation. The typology helps qualify the network externality argument for the efficacy of electronic markets. This argument emphasizes that the attractiveness of a market to a user firm depends on the number of other firms (potential business partners) in the market. We find, however, that in the case of electronic markets, network externality does not depend on the total number of firms participating in the market but rather on the number of expert (and to a lesser extent, exploration) firms in the market. Thus, the passive participants must be discounted.

We also develop the construct of legitimacy motives, which, together with efficiency motives, provides a holistic theorization of organizational motivations. In addition, we contribute to the generalizability of previous case-based research on electronic markets (e.g., Hess and Kemerer 1994) by using organizational-level survey data to test our model and a representative electronic market. Similar to most electronic markets, Polygon facilitates negotiations on product specifications and terms of exchange. Polygon strives to foster a community atmosphere and is contemplating providing ratings of participating firms that would be based on their participation history. Thus, on the basis of the characteristics and scope of Polygon, we contend that our results are generalizable to other similar electronic markets.

Limitations and Further Research

Consistent with most survey research, our results are constrained by issues related to common method variance, though we tried to minimize it in two ways: (1) We assessed nonresponse bias by comparing sample characteristics with those of the population, and (2) we used secondary data to operationalize age-based learning. To our knowledge, this is the first attempt to examine organizational participation in electronic markets empirically; as a result, readers should be cautious in generalizing these results. Replications, especially in markets in which market makers adopt business models different from those of Polygon, are needed. Additional research, for example, might study firm participation in electronic markets in which a transaction fee is the primary revenue source for the market maker (e.g., e-STEEL). In electronic markets that rely on transaction fees, participation firms that do not engage in transactions do not incur any costs (other than the initial setup costs). In such cases, we believe that participating firms are even more likely to move into the passive state, and the challenges for market makers to induce active participation would be even greater. Other business models of market makers (e.g., those based on transaction or subscription fees), other characteristics of market makers (e.g., the ownership structure-consortia-led versus third-party), strategy of the market maker (e.g., exclusive markets in which entry is contingent on some qualification versus open markets), and the structure of the industry (e.g., degree of fragmentation) also should be studied.

Further research on business-to-business electronic markets should proceed in at least two directions. First, research should study user firms in electronic markets further. Researchers need to understand the role of user firm capabilities, such as market orientation and strategic adaptability, and how these important strategic variables might change in the context of electronic markets. Researchers should examine the consequences of the nature of organizational participation, such as its influence on building interfirm relationships and firm performance. In addition, researchers should develop an understanding of the drivers of firm performance and typologies of firm strategies in electronic markets. Finally, all electronic markets are international, and understanding the role of country-specific institutional environments and the implications of omnipresent global competition among user firms is critical in developing a full appreciation of the impact of electronic markets.

Second, we investigate only one electronic market and therefore cannot contribute with regard to the role of market makers in other contexts. Further research should work toward developing a taxonomy of business-to-business electronic markets, competition among electronic markets, role of market makers and their influence on the type and characteristics of user firms, the scope and context of electronic markets (e.g., generic versus industry-specific electronic markets), and international issues such as the country of origin of market makers, among others. For example, an interesting research avenue would be to investigate the factors behind the emergence and initial success of consortia-led electronic markets.

Measures for the Nature of Organizational Participation

State                                   Measures

Exploration State        We have recently initiated the Polygon
                         service and are beginning to learn how
                         to do business through them.

                         We have learned a lot about the way to
                         do business on the Polygon, but there
                         is still much more to learn. Our comfort
                         level with doing business on Polygon is
                         improving with every passing day.

Expert State             We are comfortable with our Polygon
                         operations and are aware of the ins and
                         outs of these operations. Our dealings
                         on the Polygon are a regular part of our
                         business, and we think that there is not
                         much new to learn.

Passive State            We carry virtually no business through
                         Polygon but still are listed as a member
                         of Polygon Network and will continue to
                         be listed with Polygon.

                         We are seriously considering terminating
                         our Polygon operations.

                         We have terminated our relationship with
                         Polygon.

Sample Classification

                                                 Population
                                               Characteristics
Type of                      Sample             (Supplied by
Business (%)             Characteristics           Polygon)

Retailers                    59.1%                  60.0%

Wholesalers                  23.2%                  20.0%

Pawnbrokers                   9.5%                  10.0%

Miscellaneous
(e.g., manufacturers,
designers, appraisers)        8.2%                  10.0%

Legend for Chart:

A - Constructs and Items
B - Factor Loading (t-Value)

[a] One item was deleted after confirmatory analysis. It read,
"Our competitors joined Polygon."

[b] Two items, "Is experienced with IT" and "Gives high
importance to strategic use of IT," were deleted.


               A                                       B


Efficiency: We decided to subscribe to
Polygon because


We thought it would increase our                   .66 (11.0)
efficiency.

We expected it to reduce our costs                 .85 (15.4)
associated with running our business.

We thought it would streamline our                 .79 (13.9)
operations.

We believed that it would reduce the               .73 (12.3)
cost associated with transacting
business with our exchange partners.


Legitimacy: We decided to subscribe to
Polygon because[a]


It would provide legitimacy to our                 .76 (11.3)
organization.

It portrays us as a high-tech organization.        .74 (11.0)

The best in the business at the time               .50 (7.4)
were doing so.


Effort-Based Learning: We are interested
in your experiences with using the
Polygon system. To what extent do you
agree with the following statements?


Our personnel have spent a lot of time             .52 (5.8)
and effort learning specific techniques
used in the Polygon.

Our organization has expended a lot of             .90 (7.0)
time and effort to develop our Polygon
operations.


IT Capabilities: The following questions
pertain to information technology (IT)
capabilities for your organization.
Your firm currently[b]


Has strong IT planning capabilities.               .83 (14.7)

Has strong technical support staff.                .82 (14.5)

Has an understanding of possible benefits          .63 (10.2)
of IT applications.

Has adequate knowledge about information           .68 (11.4)
technology.


Environmental Dynamism: Please focus
on the BUYING/SELLING environment
through the Polygon. To what extent do
you agree with the following statements?


We are often puzzled by actions of                 .78 (11.5)
retailers and wholesalers.

We are often astonished by actions of              .71 (10.6)
our competitors.

We are often surprised by our customers'           .62 (9.2)
actions.


χ² (d.f., p-value)                            96.48 (94, .41)


Root mean square error of approximation            .010 (.99)
(p-value)


Goodness-of-fit index                              .96


Adjusted goodness-of-fit index                     .94


Nonnormed fit index                                .99


Parsimony normed fit index                         .73


Comparative fit index                              .99

Descriptive Statistics, Reliabilities, Average Variance Extracted, and Intercorrelations Among the Refined Measures

A= Construct
B= Mean
C= Standard Deviation
D= Reliability
E= Average Variance Extracted
F= LEGIT
G= IT-CAP
H= AGE
I= EFFORT
J= ENV-DYN

A         B         C        D        E         F

          G         H        I        J


EFF       3.400     .867     .85      .58       .343*

          .026      .069     -.166*   .092


LEGIT     2.768     .868     .71      .46

          .034      .043     -.288*   .084


IT-CAP    3.568     .798     .83

          .56       .095     -.088    -.205*


AGE       4.223     6.465

                             -.006     .084


EFFORT    3.342     .801     .69 .

54 .023
55


ENV-DYN   3.106     .718     .75       .50

*p < .01

Estimation Results from the Multinomial Logit Model

                          Coefficient         Coefficient
Independent Variable     (Expert State)a    (Passive State)a

Constant(B0)                15.674***           10.528**
                            (6.530)             (5.872)

EFF (b1)                    -3.612***           -1.858**
                            (1.267)             (.967)

LEGIT (b2)                  -.355*              -.203
                            (.221)              (.208)

AGE (b3)                    .473***             .074
                            (.191)              (.074)

AGE  AGE (b4)               -.024**             .001
                            (.013)              (.002)

EFFORT(b5)                  .389*               .293*
                            (.239)              (.213)

IT-CAP (b6)                 -4.508**            -3.836**
                           (2.252)              (2.040)

IT-CAP  IT-CAP (b7)         -.108               .247*
                            (.253)              (.182)

ENV-DYN (b8)                -2.294**            -.794
                            (1.371)             (1.048)

EFF  IT-CAP (b9)            .879***             .294
                            (.330)              (.244)

ENV-DYN  IT-CAP (b10)       .757**               .371
                            (.379)               (.292)


Log-likelihood                      -234.400

Restricted (slope = 0)
log-likelihood                      -263.425

X2                         58.06*** (d.f. = 20, p < .01)


                        Predictive Validity (n = 248)

Maximum chance criteria b                             .456

Proportional chance criteria b                        .358

Proportion correctly classified (model)               .540

A= Standard errors are in parentheses (one-tailed tests).

B= We extended Morrison's (1969) proportional chance criteria and
maximum chance criteria from a two-category discriminant problem
to a three-category problem. If a, b, and g are the proportion of
respondents in each of the three categories, maximum chance
criteria would be the maximum of these three proportions, and
proportional chance criteria would be a2 + b2 + g2.

*p < .10.

**p < .05.

***p < .01

Results for Interaction Effectsa Hypothesis Lowb Mediumb Highb Effect on the Expert State Coefficient of efficiency as a Function of IT capabilities. -1.178 -.477** .226 (.407) (.245) (.305) Coefficient of environmental dynamism as a function of IT capabilities. -.197 .408* 1.012*** (.395) (.260) (.402) Effect on the Passive State Coefficient of efficiency as a function of IT capabilities. -1.043 -.808 -.573*** (.339) (.219) (.240) Coefficient of environmental dynamism as a function of IT capabilities. .232 .529*** .824*** (.318) (.239) (.348) a Standard errors are in parentheses (two-tailed tests). b We define high as m (mean) + s (standard deviation), medium as m, and low as m - s. We now illustrate one case of the Wald test. Consider the influence of efficiency at a high level of IT capabilities. The mean for IT capabilities is 3.568, and the standard deviation is .798, which implies that high IT capabilities equal 4.366 (see Table 4). The effect of efficiency on a firm in the expert state is given as ORG - PAR = b1 + b9 IT-CAP, and at a high level of IT capabilities this effect equals ORG - PAR = b1 + b9< 4.366. Because b1= -3.612 and b = .879 (see Table 5), ORG-PAR = .226 (Table 6). Similarly, we can calculate the standard error as b12 + b9 2 4.366 2 + 2 4.366 b1 bb9 Cov(b1, b9). *p < .12 **p < .10 ***p < .05 +p < .01

REFERENCES

Achrol, Ravi Singh (1991), "Evolution of Marketing Organization: New Forms for Turbulent Environments," Journal of Marketing, 55 (October), 77-93.

--- and Louis W. Stern (1988), "Environmental Determinants of Decision Making Uncertainty in Marketing Channels," Journal of Marketing Research, 25 (February), 36-50.

Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimensions of Consumer Expertise," Journal of Consumer Research, 13 (March), 411-54.

---, John Lynch, Barton Weitz, Chris Janiszewski, Richard Lutz, Alan Sawyer, and Stacy Wood (1997), "Interactive Home Shopping: Consumer, Retailer, and Manufacturer Incentives to Participate in Electronic Marketplaces," Journal of Marketing, 61 (July), 38-53.

Anderson, James C. and David W. Gerbing (1988), "Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach," Psychological Bulletin, 103 (3), 411-23.

Armstrong, J. Scott and Terry S. Overton (1977), "Estimating Nonresponse Bias in Mail Surveys," Journal of Marketing Research, 14 (August), 396-402.

Auer, Timo and Tapio Reponen (1997), "Information System Strategy Formation Embedded into a Continuous Organizational Learning Process," Information Resource Management Journal, 10 (Spring), 32-43.

Bagozzi, Richard P. and Youjae Yi (1988), "On the Evaluation of Structural Equation Models," Journal of the Academy of Marketing Science, 16 (Spring), 74-94.

Bakos, Yannis (1991), "A Strategic Analysis of Electronic Marketplace," MIS Quarterly, 15 (September), 295-310.

--- (1998), "The Emerging Role of Electronic Marketplaces on the Internet," Communication of the ACM, 41 (August), 35-42.

--- and Erik Brynjolfsson (1993), "Information Technology, Incentives, and Optimal Number of Suppliers," Journal of Management Information Systems, 10 (Fall), 37-53.

Bass, Frank M. (1995), "Empirical Generalization and Marketing Science: A Personal View," Marketing Science, 14 (3), G6-G19.

Baum, Joel A.C. and Christine Oliver (1992), "Towards an Institutional Ecology of Organizational Founding," Academy of Management Journal, 39 (October), 1378-427.

Blattberg, Robert C., Rashi Glazer, and John D.C. Little (1994), The Marketing Information Revolution. Boston: Harvard Business School Press.

Boulding, William and Richard Staelin (1995), "Identifying Generalizable Effects of Strategic Actions on Firm Performance: The Case of Demand-Side Returns on R&D Spending," Marketing Science, 14 (Special Issue), G222-G236.

Brittain, J. (1989), "Strategy and Time Dependent Failure Probabilities: Experience, Obsolescence, and Strategic Change," Academy of Management Proceedings, 173-77.

BusinessWeek (1998), "How Electronic Commerce Could Create the Most Efficient Market of Them All: Good-Bye to Fixed Pricing?" (May 4), 71-84.

--- (1999), "A Business-to-Business E-Boom," (October 25), 62.

--- (2000), "B2B: The Hottest Net Bet Yet?" (January 17), 36-37.

Buzzell, Robert D. (1985), Marketing in an Electronic Age. Boston: Harvard Business School Press.

Chandrashekaran, Murali, Raj Mehta, Rajesh Chandrashekaran, and Rajdeep Grewal (1999), "Market Motives, Distinctive Capabilities, and Domestic Inertia: A Hybrid Model of Innovation Generation," Journal of Marketing Research, 34 (February), 95-112.

Choudhury, Vivek, Kathleen S. Hartzel, and Benn R. Konsynski (1998), "Uses and Consequences of Electronic Markets: An Empirical Examination into the Aircraft Parts Industry," MIS Quarterly, 22 (December), 471-507.

Churchill, Gilbert A., Jr. (1979), "A Paradigm for Developing Better Measures of Marketing Constructs," Journal of Marketing Research, 16 (February), 64-73.

Copeland, Duncan G. and James L. McKenney (1988), "Airline Reservation Systems: Lessons from History," MIS Quarterly, 12 (September), 353-70.

Curry, David J. (1993), The New Marketing Research System: How to Use Strategic Database Information for Better Marketing Decisions. New York: John Wiley & Sons.

Daft, Richard L. and Robert H. Lengel (1986), "Organization Information Requirements, Media Richness, and Structural Design," Management Science, 32 (May), 554-71.

Day, George S. (1994), "The Capabilities of Market-Driven Organizations," Journal of Marketing, 58 (October), 37-52.

--- and Rashi Glazer (1994), "Harnessing the Marketing Information Revolution: Towards the Market-Driven Learning Organization," in The Marketing Information Revolution, R.C. Blattberg, R. Glazer, and J.D.C. Little, eds. Boston: Harvard Business School Press, 270-88.

--- and Prakash Nedungadi (1994), "Managerial Representation of Competitive Advantage," Journal of Marketing, 58 (April), 31-44.

--- and Robin Wensley (1988), "Assessing Advantage: A Framework for Diagnosing Competitive Superiority," Journal of Marketing, 52 (April), 1-20.

Deephouse, David L. (1996), "Does Isomorphism Legitimate?" Academy of Management Journal, 39 (August), 1024-1039.

Deshpande, Rohit and Frederick E. Webster (1989), "Organizational Culture and Marketing: Defining a Research Agenda," Journal of Marketing, 53 (January), 3-15.

Dess, Gregory G. and Donald W. Beard (1984), "Dimensions of Organizational Environments," Administrative Science Quarterly, 29 (March), 52-73.

Dickson, Peter R. (1992), "Toward a General Theory of Competitive Rationality," Journal of Marketing, 56 (January), 69-83.

DiMaggio, Paul J. and Walter W. Powell (1983), "The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in Organizational Behavior," American Sociological Review, 48 (April), 147-60.

The Economist (2000), "Seller Beware," (March 4), 61-62.

Ewusi-Mensah, Kweku and Zbigniew H. Przasnyski (1991), "On Information System Project Abandonment: An Exploratory Study of Organizational Practices," MIS Quarterly, 15 (March), 67-88.

Fiol, C. Marlene and Marjorie A. Lyles (1985), "Organizational Learning," Academy of Management Review, 10 (October), 803-13.

Fornell, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error," Journal of Marketing Research, 28 (February), 39-50.

Frazier, Gary L. (1983), "Interorganizational Exchange Behavior in Marketing Channels: A Broadened Perspective," Journal of Marketing, 47 (Fall), 68-78.

Galbraith, Jay R. (1973), Designing Complex Organizations. Reading, MA: Addison-Wesley.

Garvin, David A. (1993), "Building a Learning Organization," Harvard Business Review, 71 (July/August), 78-91.

Ginzberg, Michael J. (1981), "Key Recurrent Issues in MIS Implementation Process," MIS Quarterly, 5 (June), 47-59.

Glazer, Rashi (1991), "Marketing in an Information Intensive Environment: Strategic Implications of Knowledge as an Asset," Journal of Marketing, 55 (October), 1-19.

--- and Allen M. Weiss (1993), "Marketing in Turbulent Environments: Decision Processes and Time Sensitivity of Information," Journal of Marketing Research, 30 (November), 509-21.

Golder, Peter N. and Gerard J. Tellis (1993), "Pioneer Advantage: Marketing Logic or Marketing Legend?" Journal of Marketing Research, 30 (August), 158-70.

Guha, Subo, Varun Grover, William J. Kettinger, and James T.C. Teng (1997), "Business Process Change and Organizational Performance: Exploring an Antecedent Model," Journal of Management Information Systems, 14 (1), 119-54.

Gurbaxani, Vijay and Seungjin Whang (1991), "The Impacts of Information Systems on Organizations and Markets," Communications of the ACM, 34 (January), 59-73.

Hair, Joseph F., Rolph E. Anderson, Ronald L. Tatham, and William C. Black (1995), Multivariate Data Analysis. Englewood Cliffs, NJ: Prentice Hall.

Haunschild, Pamela R. (1993) "Interorganizational Imitation: The Impact of Interlocks on Corporate Acquisition Activity," Administrative Science Quarterly, 38 (December), 564-92.

--- and Anne S. Miner (1997), "Modes of Interorganizational Imitation: The Effects of Outcome Salience and Uncertainty," Administrative Science Quarterly, 42 (September), 472-500.

Haverman, Heather (1993), "Follow the Leader: Mimetic Isomorphism and Entry into New Markets," Administrative Science Quarterly, 38 (4), 593-627.

Heide, Jan B. and Allen M. Weiss (1995), "Vendor Consideration and Switching Behavior in High-Technology Markets," Journal of Marketing, 59 (July), 30-43.

Hess, C.M. and C.F. Kemerer (1994), "Computerized Loan Organization Systems: An Industry Case Study of the Electronic Markets Hypothesis," MIS Quarterly, 18 (September), 251-74.

Hoffman, Donna L. and Thomas P. Novak (1996), "Marketing in Hypermedia Computer-Mediated Environments: Conceptual Foundations," Journal of Marketing, 60 (July), 50-68.

Huber, George P. (1991), "Organizational Learning: The Contributing Processes and the Literatures," Organization Science, 2 (February), 88-115.

Jaccard, James, Robert Turrisi, and Choi K. Wan (1990), Interaction Effects in Multiple Regression. Newbury Park, CA: Sage Publications.

Kaplan, Steven and Mohanbir Sawhney (2000), "E-Hubs: The New B2B Marketplace," Harvard Business Review, (May/June), 97-103.

King, William R. and Thompson S.H. Teo (1996), "Key Dimensions of Facilitators and Inhibitors for the Strategic Use of Information Technology," Journal of Management Information Systems, 12 (Spring), 35-53.

Klein, Lisa R. and John A. Quelch (1997), "Business-to-Business Market Making on the Internet," International Marketing Review, 14 (5), 345-61.

Klein, Saul, Gary L. Frazier, and Victor J. Roth (1990), "A Transaction Cost Analysis Model of Channel Integration in International Markets," Journal of Marketing Research, 27 (May), 196-208.

Krantz, Michael (1999), "The Next E-volution," Time, (July 12), 47.

Lacity, Mary C. and Rudy Hirschheim (1993), "The Information Systems Outsourcing Bandwagon," Sloan Management Review, 35 (Fall), 73-86.

Lee, Ho Geun and Theodore H. Clark (1997), "Market Process Reengineering Through Electronic Market Systems: Opportunities and Challenges," Journal of Management Information Systems, 13 (Winter), 113-36.

Levinthal, Daniel A. and James G. March (1993), "The Myopia of Learning," Strategic Management Journal, 14, 95-112.

MacInnis, Deborah J., Christine Moorman, and Bernard J. Jaworski (1991), "Enhancing and Measuring Consumers' Motivation, Opportunity, and Ability to Process Brand Information from Ads," Journal of Consumer Research, 55 (October), 32-53.

Malone, Thomas W., Jo Anne Yates, and Robert I. Benjamin (1994), "Electronic Markets and Electronic Hierarchies," in Information Technology and the Corporation of the 1990s, Thomas J. Allen and Michael S. Scott Morton, eds. New York: Oxford University Press, 61-83.

March, James G. (1991), "Exploration and Exploitation in Learning Organization," Organization Science, 2 (February), 71-87.

Mata, Francisco J., William L. Fuerst, and Jay B. Barney (1995), "Information Technology and Sustained Competitive Advantage: A Resource-Based View," MIS Quarterly, 19 (December), 487-505.

McGee, James and Benn Konsynski (1989), "A Note on Electronic Data Interchange," Harvard Business School, Case No. 9-190-022.

Mehta, Raj, Rajdeep Grewal, and Eugene Sivadas (1996), "International Direct Marketing on the Internet: Do Internet Users Form a Global Segment?" Journal of Direct Marketing, 10 (Winter), 45-58.

Merton, Robert K. (1957), Social Theory and Social Structure. Glencoe, IL: The Free Press.

Meyer, Alan D. and James B. Goes (1988), "Organizational Assimilation of Innovations: A Multilevel Contextual Analysis," Academy of Management Journal, 31 (4), 897-923.

Meyer, John W. and Brian Rowan (1977), "Institutionalized Organizations: Formal Structure as Myth and Ceremony," American Journal of Sociology, 83 (September), 340-63.

Mody, Ashoka (1989), "Firm Strategies for Costly Engineering Learning," Management Science, 35 (4), 496-512.

Moorman, Christine, Gerald Zaltman, and Rohit Deshpande (1992), "Relationship Between Providers and Users of Marketing Research: The Dynamics of Trust Within and Between Organizations," Journal of Marketing Research, 29 (August), 314-28.

Morrison, Donald G. (1969), "On the Interpretation of Discriminant Analysis," Journal of Marketing Research, 6 (May), 156-63.

Myers, John G., William G. Massy, and Stephen A. Greyser (1980), Marketing Research and Knowledge Development. Englewood Cliffs, NJ: Prentice Hall.

Nelson, Richard R. and Sidney G. Winter (1982), An Evolutionary Theory of Economic Change. Cambridge, MA: Harvard University Press.

Newman, Michael and Rajiv Sabherwal (1996), "Determinants of Commitment to Information Systems Development: A Longitudinal Investigation," MIS Quarterly, 20 (March), 23-54.

Nunnally, Jum C. and Ira H. Bernstein (1994), Psychometric Theory. New York: McGraw-Hill.

O'Callaghan, Ramon, Patrick J. Kaufmann, and Benn R. Konsynski (1992), "Adoption Correlates and Share Effect of Electronic Data Interchange Systems in Marketing Channels," Journal of Marketing, 56 (April), 45-56.

Oliver, Christine (1990), "Determinants of Interorganizational Relationships: Integration and Future Direction," Academy of Management Review, 15 (2), 241-65.

O'Reilly, Charles A., III, and Jennifer A. Chatman (1994), "Working Smarter and Harder: A Longitudinal Study of Managerial Success," Administrative Science Quarterly, 39 (December), 603-27.

Osborn, Richard N. and John Hagedoorn (1997), "The Institutionalization and Evolutionary Dynamics of Interorganizational Alliances and Networks," Academy of Management Journal, 40 (2), 261-78.

Pangarkar, Nitin and Saul Klein (1998), "Bandwagon Pressures and Interfirm Alliances in the Global Pharmaceutical Industry," Journal of International Marketing, 6 (2), 54-73.

Parasuraman, A. and Dhruv Grewal (2000), "Serving Customers and Consumers Effectively in the Twenty-First Century: A Conceptual Framework and Overview," Journal of the Academy of Marketing Science, 28 (1), 9-16.

Peterson, Robert A., Sridhar Balasubramanian, and Bart J. Bronnenberg (1997), "Exploring the Implications of Internet for Consumer Marketing," Journal of the Academy of Marketing Science, 25 (4), 329-46.

Quelch, John A. and Lisa R. Klein (1996), "The Internet and International Marketing," Sloan Management Review, 37 (Spring), 60-75.

Reich, Blaize H. and Izak Benbasat (1990), "An Empirical Investigation of Factors Influencing the Success of Customer-Oriented Strategic Systems," Information Systems Research, 1 (September), 325-47.

Rindfleisch, Aric and Jan B. Heide (1997), "Transaction Cost Analysis: Past, Present, and Future Applications," Journal of Marketing, 61 (October), 30-54.

Roberts, Peter W. and Royston Greenwood (1997), "Integrating Transaction Costs and Institutional Theories," Academy of Management Review, 22 (April), 346-73.

Schulz, Martin (1998), "Limits to Bureaucratic Growth: The Density Dependence of Organizational Rule Births," Administrative Science Quarterly, 43 (December), 845-76.

Scott, W. Richard (1987), "The Adolescence of Institutional Theory," Administrative Science Quarterly, 32 (December), 493-511.

Simon, Herbert A. (1991), "Bounded Rationality and Organizational Learning," Organization Science, 2 (February), 125-34.

Sinkula, James M. (1994), "Market Information Processing and Organizational Learning," Journal of Marketing, 58 (January), 35-45.

Slater, Stanley F. and John C. Narver (1995), "Market Orientation and the Learning Organization," Journal of Marketing, 59 (July), 63-74.

Stern, Louis W. and Patrick J. Kaufmann (1985), "Electronic Data Interchange in Selected Consumer Industries: An Interorganizational Perspective," in Marketing in an Electronic Age, R. Buzzell, ed. Boston: Harvard Business School Press, 52-74.

Suchman, Mark C. (1995), "Managing Legitimacy: Strategic and Institutional Approaches," Academy of Management Review, 20 (July), 571-610.

Tushman, Michael L. and Philip Anderson (1997), Managing Strategic Innovation and Change. New York: Oxford University Press.

and David A. Nadler (1978), "Information Processing as an Integrating Concept in Organizational Design," Academy of Management Review, 3 (July), 613-24.

Weiss, Allen M. and Jan B. Heide (1993), "The Nature of Organizational Search in High Technology Markets," Journal of Marketing Research, 30 (May), 220-33.

Williamson, Oliver E. (1987), "Transaction Cost Economics: Comparative Contracting Perspective," Journal of Economic Behavior and Organization, 8 (4), 617-25.

(1992), "Markets, Hierarchies, and the Modern Corporation: An Unfolding Perspective," Journal of Economic Behavior and Organization, 17 (3), 335-52.

GRAPH: FIGURE 1: Conceptual Framework

GRAPH: FIGURE 2: Influence of Age-Based Learning

~~~~~~~~

By Rajdeep Grewal; James M. Comer and Raj Mehta

Rajdeep Grewal is Assistant Professor of Marketing, Smeal College of Business, Pennsylvania State University.

James M. Comer is Professor of Marketing

Raj Mehta is Associate Professor of Marketing, University of Cincinnati

Antecedents and Consequences of Merit Pay Fairness for Industrial Salespeople 

How do salespeople make judgments of merit pay fairness? By what mechanisms do fairness judgments influence the performance and commitment of salespeople? Using equity and social exchange theories, the authors examine these questions for industrial salespeople who work in a Fortune 500 firm and provide four key findings. First, of the three dimensions of fairness judgments, they find the interactional fairness dimension to be relatively more important than procedural or distributive fairness in influencing job outcomes of salespeople. Second, supervisory behaviors have significant influence in shaping salespeople's fairness judgments, particularly judgments of distributive and interactional fairness. Third, the results underscore the contrasting mediating role of trust in supervisor and job satisfaction. Although trust in the supervisor is important in reducing salespeople's opportunistic behaviors, the authors find job satisfaction to be important in enhancing their loyalty to the organization. Fourth, salespeople's job performance is influenced directly by extrinsic factors such as fairness of current rewards and potential for rewards. In addition, the authors outline implications for theory and practice.

Of the various decisions made by sales managers, merit pay-raise decisions (i.e., decisions about incremental pay awarded to employees on the basis of their performance) are arguably one of the most important. Merit pay raises are instrumental in directing salespeople's behaviors toward organizational goals and in facilitating retention of high-performing salespeople (Bartol 1999; Lawler 1990). Retention is facilitated by organizations differentially recognizing salespeople whose performance is exemplary. Merit pay raises also have a compounding effect on a salesperson's financial well-being. A gap of a few percentage points in merit pay-raise amounts can accumulate over a salesperson's career to produce a significant financial impact on that person's time-value of employment income. Studies consistently show that salespeople value pay raises more than any other performance reward, including promotion opportunities, fringe benefits, and recognition awards (Chonko, Tanner, and Weeks 1992; Churchill, Ford, and Walker 1979; Cron, Dubinsky, and Michaels 1988; Ford, Churchill, and Walker 1985; Ingram and Bellenger 1983; Money and Graham 1999).

Despite the recognition of pay valence for employees in general and salespeople in particular, dissatisfaction with pay and compensation plans remains prominent in employee surveys (Denton 1991; Farnham 1989; Leonard 2001). When pay expectations are not met, employees may believe that the organization has violated its obligations and disregarded its commitments (Lester et al. 2002). However, this does not mean that salespeople expect to receive the highest monetary reward; rather, they expect a fair level of reward relative to their performance (Denton 1991; Livingstone, Roberts, and Chonko 1995). Thus, if every salesperson received the same reward regardless of performance, it not only would raise issues of inequity and distress but also would likely undermine salespeople's motivation to raise their effort and performance level (Denton 1991).

In contrast, fair treatment of employees sets into motion a social exchange process by which the supervisory and organizational efforts to make fair decisions engender an obligation to reciprocate on the part of employees. Although previous research has provided general support for this reciprocity effect (Aryee, Budhwar, and Chen 2002), our specific focus is on both the antecedents and the consequences of fairness in salesperson pay decisions. In particular, we examine how perceptions of fair pay decisions not only strengthen salespeople's long-term relationship and attachment to their organization but also encourage reciprocity with functional behaviors, including less opportunism (Ramaswami, Srinivasan, and Gorton 1997) and greater job performance, which enables enhanced organizational effectiveness (Netemeyer et al. 1997). Moreover, we examine how salespeople formulate pay fairness perceptions as a result of supervisor and management actions.

Although previous research has explicated the social exchange basis of employee attitudes and behaviors, notable gaps remain in the application of the equity and social exchange theory frameworks to an understanding of the antecedents and consequences of salespeople's perceptions of merit pay-raise fairness. First, although research has identified several facets of fairness (namely, distributive, procedural, and interactional), little work has been done to examine the differential effect of these dimensions empirically in a single study (Konovsky and Pugh 1994; Kumar, Scheer, and Steenkamp 1995; Livingstone, Roberts, and Chonko 1995; Netemeyer et al. 1997; Organ 1988). Some researchers argue that the distributive dimension is most important as it relates to the magnitude of reward (e.g., Organ 1988). In contrast, in nonsales contexts, Konovsky and Pugh (1994) and others suggest that procedural fairness is more significant as it pertains to processes that determine rewards. Most studies ignore interactional fairness, that is, how salespeople are treated during the rewards process. The lack of attention to interactional fairness is notable, because issues pertaining to human dignity and respect are assuming greater importance in the workplace. Thus, current knowledge about the relative significance of distributive, procedural, and interactional fairness judgments in the context of merit pay-raise decisions is largely deficient.

Second, recent studies in social exchange research suggest that job satisfaction (Netemeyer et al. 1997) and supervisory trust (Aryee, Budhwar, and Chen 2002; Konovsky and Pugh 1994) mediate the influence of fairness dimensions on outcomes. Unfortunately, empirical studies in sales management have largely failed to study the presence and importance of such mediating processes. We were unsuccessful in finding a single study that examines different, potentially competing mediating processes. Netemeyer and colleagues (1997, p. 95) echo this gap in their observation that "sales-oriented ... research ... should consider several mediators simultaneously to determine which ones have stronger effects." Addressing this gap will enable integration of processes that heretofore have been examined independently.

Third, in studies of fairness consequences, sales researchers have focused their attention primarily on social exchange concepts, such as commitment and trust. Little research exists on the effects of fairness on job performance or opportunistic behaviors of salespeople. Performance on assigned tasks and, in the context of salespeople, opportunistic behaviors (e.g., "managing" data and effort to create favorable impressions and evaluations) likely have a significant effect on the bottom line. Understanding how fairness in merit pay raises can influence such bottom-line consequences holds implications for managers and researchers alike.

Finally, in general, sales researchers have neglected the antecedents of fairness. This neglect is notable because studies of fairness consequences do not tell a complete story. They establish that fairness perceptions matter, but they provide little insight into what managers can do to promote fairness. As such, the focus on fairness consequences in prior studies is likely to be less useful in drawing managerial implications.

We aim to address the preceding gaps by ( 1) examining three distinct fairness dimensions, ( 2) exploring the mediating mechanisms that govern fairness effects by using a nomological model rooted in social exchange and equity theories, ( 3) examining outcomes (commitment, performance, and opportunism) that are important for organizational effectiveness, and ( 4) drawing on diverse literature to propose and test an initial set of fairness antecedents. Overall, inclusion of all three fairness dimensions avoids misspecification bias due to omitted constructs, and an understanding of the key, competing mediating pathways is likely to yield theoretical payoffs and more concrete managerial guidelines for enhancing fairness-outcome relationships.

Merit Pay Fairness: Conceptualization and Framework

Fairness issues in organizations span several facets of work, including job design, performance evaluation, monetary rewards, and resource allocation. The focus of this study is on fairness of merit pay-raise decisions. Although the literature on pay fairness for salespeople is sparse, three points underscore the significance of research in this area. First, merit pay systems serve an instrumental function by directing the individual salesperson's behaviors toward fulfilling organizationally mediated sales goals and toward linking rewards received to the achievement of sales goals. Second, merit pay systems facilitate greater work motivation by differentially rewarding top performers over marginal performers. Studies have shown that a discriminating pay system can increase employees' motivation to perform by as much as 40% (Lawler 1990). Third, merit pay systems play a crucial role in retaining more "effective" salespeople. Opportunities (or lack thereof) for enhanced compensation levels influence salespeople's decisions to quit and their commitment to the organization (Lawler 1990; Singh, Verbeke, and Rhoads 1996).

Figure 1 presents the framework we used to study the antecedents and consequences of fairness dimensions based on equity and social exchange theories. Our focus is on fairness dimensions (i.e., understanding how employees determine whether they have been treated fairly in their merit pay-raise decisions) and fairness processes (i.e., examining the nature and strength of relationships involving antecedents and consequences). We begin our discussion with fairness dimensions.

Distributive, Procedural, and Interactional Dimensions of Merit Pay Fairness

Distributive fairness involves magnitude of rewards. Salespeople gauge the reward magnitude relative to their input and then compare this ratio with reward-to-input ratios of similar employees (Livingstone, Roberts, and Chonko 1995). People perceive inequity if their reward-to-input ratios do not compare favorably with such ratios of others. However, because employees may not have complete information about the reward-to-input ratio of others, they sometimes derive distributive fairness evaluations by comparing the rewards received with their own expectations (Netemeyer et al. 1997).

Procedural fairness deals with how decisions are made. The degree of individual control over the decision process shapes employees' views about procedural fairness (Lind and Tyler 1988). Researchers have identified two types of controls: ( 1) process control, or the degree of influence afforded to employees over the procedures that supervisors use to arrive at pay raise decisions, and ( 2) decision control, or the amount of direct control employees have in determining pay-raise outcomes.

Finally, interactional fairness involves how decisions are communicated. As such, it focuses on the social enactment of procedures and the quality of supervisor-salesperson interaction during the pay decision process (Goodwin and Ross 1992). Often, noninstrumental values, such as dignity, respect, and social standing, are evoked in such evaluations (Folger and Konovsky 1989).

Fairness Consequences: Direct and Mediated Pathways

Following Katz and Kahn (1978), we focus on three organizational outcomes: attachment, performance, and opportunism. Katz and Kahn prescribe that ( 1) when employees enter the organization, they should be engaged with it (organizational commitment); ( 2) as employees, they should carry out their roles in a dependable, superior fashion (job performance); and ( 3) employees should not engage in activities that thwart or undermine organizational effectiveness (opportunism). As we noted previously, fairness studies have provided greater attention to the organizational commitment outcome and less attention to the other two outcomes.

Two distinct pathways, direct and mediated, have been proposed for the influence of fairness dimensions on outcomes. The direct pathway hypothesizes that distributive, procedural, and interactional fairness have a direct impact on outcomes (George 1991; Konovsky and Cropanzano 1991). Notably, the theoretical mechanism for the direct hypothesis is the notion of reciprocity in a balance theory framework. For example, Greenberg (1990) finds that employees who experience pay inequity respond with acts of deviance or opportunism to address the inequity. In contrast, fair employer actions are reciprocated with favorable employee responses, including greater performance and commitment.

The mediating pathway proposes that the influence of fairness dimensions on employee outcomes is mediated by two social exchange variables: supervisor trust and job satisfaction. Perceptions of fairness are likely to promote enhanced feelings of job satisfaction (because of the attainment of valued rewards) and trust in the supervisor (for making a good faith effort to be fair) (Konovsky and Pugh 1994; Roberson, Moye, and Locke 1999). In turn, job satisfaction and trust in supervisors are posited to influence employees' commitment, performance, and opportunism. Although prior studies have examined the mediating role of either job satisfaction (Netemeyer et al. 1997) or supervisor trust (Konovsky and Pugh 1994), few studies have examined the simultaneous and possibly differential mediating roles of job satisfaction and trust in fairness decisions. To the extent that these competing and differential mediating effects are supported, our study integrates prior studies and advances the understanding of mechanisms that underpin the relationships between fairness and critical work outcomes. Next, we develop hypotheses separately for each chain in the mediation pathway.

Influence of Fairness Dimensions on Supervisor Trust and Job Satisfaction

Because trust reflects salespeople's willingness to rely on their supervisors to protect their interests, it is a key element in the development and maintenance of social exchange relationships (Morgan and Hunt 1994). Why would salespeople trust their supervisors? Existing research indicates that an important driver of trust is employees' perceptions of procedural fairness (Dwyer, Schurr, and Oh 1987; Flaherty and Pappas 2000; Konovsky and Pugh 1994; Lind and Tyler 1988). Konovsky and Cropanzano (1991) suggest that the use of fair procedures generates expectations of fair treatment in the long run. In turn, these expectations lead to a generalized sense of positive regard for (and trust in) the supervisor who uses fair procedures. In contrast, supervisors who have a reputation for unfair, unilateral action signal that they are only interested in their own welfare and organizational welfare. Such a negative reputation is likely to reduce employees' trust in the supervisor (Ganesan 1994). Furthermore, when supervisors demonstrate respect for the rights and dignity of salespeople through communication and high-quality interactions, they signal that salespeople are valued members of the group (Folger and Konovsky 1989). Treatment of salespeople in a manner that reinforces their self-worth helps enhance their trust in the supervisor (Lind and Tyler 1988)

Two opposing arguments can be posited for the relationship between distributive fairness and supervisor trust. Konovsky and Pugh (1994, p. 658) suggest that "a norm of distributive fairness implies that the parties to an exchange give benefits with the expectation of receiving comparable benefits in the short run." In other words, fair compensation is expected and therefore will not contribute to increased trust. In support of this, Konovsky and Pugh find no empirical association between distributive fairness and supervisor trust. Alternatively, taking a long-term view of salespeople's relationship with the organization, salespeople's distributive fairness perceptions could exhibit social exchange effects if the merit pay awards even out in the long run. We posit that the effect of distributive fairness is in line with Konovsky and Pugh's proposition, but we note the potential for an alternative explanation.

In addition, each fairness dimension is hypothesized to have an unequivocal, positive influence on salespeople's job satisfaction. Job satisfaction is defined as a pleasurable or positive emotional state that results from self-appraisal of a job or job experiences (Livingstone, Roberts, and Chonko 1995). In a merit pay context, employees who experience distributive fairness are likely to exhibit greater satisfaction (Moorman 1991; Netemeyer et al. 1997). According to equity theory, the greater the discrepancy between the amount employees believe they should receive and the actual amount they receive, the greater is their tension or dissatisfaction (Lawler 1990; Livingstone, Roberts, and Chonko 1995). Moreover, job satisfaction is likely to be positively associated with the degree to which the merit pay system adheres to salespeople's sense of procedural fairness (Roberson, Moye, and Locke 1999). For example, employees who perceive that procedures are unfair may entertain feelings that they would have obtained a higher merit pay under a procedure that was "fairer" and consequently might feel angry and dissatisfied (Folger 1986). Furthermore, employees' perceptions of interactional fairness may be associated with how salespeople perceive management's valuation of their contribution, thereby affecting job satisfaction (Moorman 1991). Although similar value judgments can be communicated through formal procedures, the quality interactions with the supervisor in pay decisions provide compelling evidence of an individual employee's worth on the job. Thus, we hypothesize the following:

H₁: The greater the procedural or interactional fairness perceptions, the greater is salespeople's trust in the supervisor.

H₂: Distributive fairness perceptions are unrelated to trust in the supervisor.

H₃: The greater the distributive, procedural, or interactional fairness perceptions, the greater is employees' job satisfaction.

The Consequences of Supervisor Trust and Job Satisfaction

It has been hypothesized that salespeople with high levels of supervisor trust perform relatively better, evidence greater commitment, and are more restrained in opportunistic behaviors than are salespeople with low levels of supervisor trust. Specifically, in terms of job performance, a salesperson with a high level of trust in the supervisor is more likely to cooperate with the latter because of a desire to make the relationship work and to add value to it (Morgan and Hunt 1994). A way to add value is for employees to enhance their job performance, because this has the effect of enhancing the supervisor's performance (Costa, Roe, and Tallieu 1991; Rich 1997; Tyagi 1985).

Trust in the supervisor also enhances employees' perceptions of the value of staying in a long-term relationship with the organization, resulting in increased commitment (Moorman, Zaltman, and Deshpandé 1992). Morgan and Hunt (1994) note that employees will commit themselves to organizations that provide trustful work relationships. Likewise, McDonald (1981, p. 834) notes that "mistrust breeds mistrust and ... would also serve to decrease commitment in the relationship." Moorman, Zaltman, and Desphandé (1992) and Morgan and Hunt provide empirical support for this linkage.

Finally, employees often behave politically and engage in intentional acts of influence to enhance or protect their self-interest (Crant and Bateman 1993). Political behaviors are opportunistic and may include selectively presenting or intentionally distorting information given to supervisors and others (Ramaswami 1996). In sales settings, opportunistic behaviors arise when salespeople fail to engage in behaviors that are warranted but not measured by the performance system and/or when they engage in irrelevant behaviors because they are measured by the performance system (Jaworski and MacInnis 1989). They may also manipulate the information processed by supervisors for setting sales goals (Ramaswami, Srinivasan, and Gorton 1997), shirk from making the required number of sales calls, or fail to fulfill promises (Ganesan, Weitz, and John 1993). However, when salespeople trust their supervisors, they are more likely to believe that they can achieve better long-term rewards by reciprocating with cooperative rather than self-interested behaviors (Anderson and Narus 1990).

In contrast, we expect job satisfaction to influence commitment and opportunistic behaviors but not job performance. Specifically, regarding job performance, a strong argument can be made that a satisfied employee is also a productive employee (Organ 1977; Petty, McGee, and Cavender 1984). However, researchers have failed to find evidence of strong association between satisfaction and performance, despite the many studies that have been conducted to uncover this relationship (Brown and Peterson 1993). In a meta-analysis, Iaffaldano and Muchinsky (1985, p. 270) suggest that satisfaction and performance form only "an illusory correlation ... between two variables that we logically think should interrelate ... [but] in fact do not." In a more recent meta-analysis, Judge and colleagues (2001) update previous findings and note that though the satisfaction-performance relationship is weak (correlation ∼ .30), it is positive and significant. Nevertheless, given this mixed evidence, we do not hypothesize a relationship between job satisfaction and performance.

When an organization provides employees with satisfying jobs, commitment levels should increase. Existing empirical models seem to provide support for job satisfaction as an antecedent to commitment (Brown and Peterson 1993; Dailey and Kirk 1992; Johnson, Barksdale, and Boles 2001; Johnston et al. 1990; Summers and Hendrix 1991). Williams and Hazer (1986, p. 230) state that "through a process of the evaluation of costs and benefits, individual needs and desires are satisfied, and the resulting affective state becomes associated with the organization.... Commitment results from this association." Martin and Bennett (1996) suggest that whereas job satisfaction is an affective response to specific work-related facets (e.g., pay), organizational commitment represents an affective response to the entire organization. They state (p. 85), "as individual needs are satisfied, the resulting satiated state becomes associated with the focal organization." For salespeople, Singh, Verbeke, and Rhoads (1996) provide empirical support for this association. Likewise, if salespeople are satisfied with their jobs, they are less likely to be opportunistic out of a sense of debt, obligation, or even decency. On the basis of the preceding discussion, we hypothesize the following:

H₄: The greater the salesperson's trust in the supervisor, the greater is his or her performance and organizational commitment and the lower is the tendency to engage in opportunistic behavior.

H₅: The greater the salesperson's job satisfaction, the greater is his or her organizational commitment and the lower is the tendency to engage in opportunistic behavior.

H₆: The salesperson's job satisfaction is unrelated to his or her job performance.

Antecedents of Distributive, Procedural, and Interactional Fairness Judgments

Previous fairness research has focused primarily on fairness consequences in an effort to explain various organizational behaviors and outcomes and has paid little attention to the antecedents of fairness perceptions. The few studies that do examine fairness antecedents have invariably grounded their work in a managerial perspective for selecting antecedent factors. We are not surprised by this, given that the study of fairness antecedents is apparently of greater interest for managerial practice than for theory-driven researchers. However, this apparent difference in focus does not pass careful scrutiny, because the study of fairness mechanisms must try to tie together its antecedents and consequences (Greenberg 1990).

Toward this end, we identify antecedent factors for each fairness dimension that ( 1) originate in either procedural justice or equity theory, ( 2) can be easily acted on by managers, and ( 3) have received empirical attention in the academic literature. However, there might be other managerially relevant antecedent factors that are not included in our study in order to keep the empirical model manageable and reasonable. Because our intent is to lay the foundation for theory building in the area of fairness antecedents, we provide this work as an initial attempt in this direction. As such, we draw propositions rather than hypotheses to capture the relationships involving fairness antecedents.

Before developing specific propositions, we elaborate on the broad theoretical mechanisms underlying the effects of modeled antecedents. Notably, distributive fairness perceptions will be influenced by factors that are related to supervisory evaluation of outcomes relative to inputs. With regard to inputs, employees want supervisors to use performance measures that are relevant and important to their jobs. With regard to outcomes, employees would like supervisors to maintain correspondence between their inputs (i.e., performance) and their outcomes (i.e., rewards). In addition, salespeople expect supervisors to apply standards consistently for everyone. Finally, salespeople may be concerned as much about future input-outcome relationships as they are about current ones. Thus, the supervisor's attempt to improve future performance-reward linkages may be important.

Procedural fairness perceptions will be influenced by process factors that directly or indirectly contribute to employees' sense of influence on the merit reward decision. The specific process factors may include ( 1) the presence of mechanisms that enable employees to provide their views and ( 2) the absence of mechanisms that signal undue influence of other employees. The first category includes factors associated with employee voice mechanisms, including participation, or the extent to which the employee has voice in setting work goals and the means to achieve them (Goodwin and Ross 1992; Korsgaard, Schweiger, and Sapienza 1995), and improvement plan, or the extent to which the supervisor works with the employee in developing a plan to improve future outcomes or rewards. The second category includes factors that level the playing field, such as the use of appropriate performance measures (Leventhal 1980) and their consistent application across all employees (Greenberg 1986).

Interactional fairness is based not on the ability to shape outcomes but on satisfaction of noninstrumental values such as social standing, dignity, and respect. When supervisors help salespeople develop a plan to improve future performance and communicate clearly that the organization is concerned for their well-being, interactional fairness is likely enhanced. In addition, supervisors promote interactional fairness when they allow the employees to participate in setting work goals.

Performance-reward linkage. We argue that only when supervisors use performance information that is reliably related to outcomes will employees be favorably disposed to the distributive fairness of their outcomes. Linking pay and performance has been a cornerstone of employee compensation (Lawler 1990). However, this ideal often is not achieved, because organizational rewards may be based on several factors beyond performance, including budget availability, political behavior, seniority, supervisor-employee dependence, and other extrarole behaviors (Bartol and Martin 1989; Podsakoff and MacKenzie 1994; Turban and Jones 1988). In addition, managers may have other reasons (e.g., equality versus equity) for not linking pay to performance (Miceli 1993; Sarin and Mahajan 2001). In a study of managers, Markham (1988) finds a weak significant correlation (r = .19) between supervisor's performance rating and the merit raise received by employees. When organizations fall short of fulfilling the core principle of linking pay to performance, distributive fairness of merit pay decisions is likely to be compromised. Thus:

P₁: The greater the perceived linkage between job performance and pay outcomes, the greater are a salesperson's perceptions of distributive fairness.

Consistent/unbiased application of performance standards. When employees observe that reasonable standards are not applied consistently across all employees, both distributive and procedural fairness judgments are likely to be affected (Bartol 1999; Kumar, Scheer, and Steenkamp 1995). In a limited budget scenario, employees know that inconsistent application of standards could upset the input-outcome relationship by providing greater allocation to some and less allocation to others. In Sashkin and Williams's (1990) study, "playing favorites" was a factor employees often mentioned when asked to describe managers' unfair actions. Moreover, when performance standards are applied inconsistently, the implication is that the process of determining pay allocations is subverted. Tyler (1989) suggests that salespeople are particular that their management creates a neutral arena (i.e., a level playing field) in which to resolve their problem or conflict. Thus:

P₂: The greater the consistent/unbiased application of standards, the greater are a salesperson's perceptions of (a) distributive fairness and (b) procedural fairness.

Performance improvement plans. Performance appraisal processes often serve two basic purposes: ( 1) They provide information for managerial decisions such as pay and promotion, and ( 2) they enable supervisors to counsel employees on ways to improve future job performance (Dubinsky and Barry 1982; Muczyk and Gable 1987). Because of the significance of the latter in improving long-term pay returns, performance improvement plans are likely to influence distributive, procedural, and interactional fairness judgments. The focus on performance improvement informs the employee that the supervisor is interested in the employee's receiving greater future rewards, and it engenders feelings of reduced unfairness about rewards received in the current period. Furthermore, a performance improvement plan enables the supervisor to provide advance notice to the employee of future expectations (Martin and Bartol 1998) and consequently can result in greater role clarity and more favorable perceptions of procedures (Gilliland 1993; Ramaswami 1996). Finally, when using performance improvement plans, supervisors provide useful how-to information about progress toward desired outcomes (Bartol 1999). This feedback signals that supervisors are concerned about employees' personal development and that they care enough to spend time and effort to improve employees' performance, thus enhancing perceptions of interactional fairness. Thus:

P₃: The greater the supervisory focus on developing plans for performance improvement, the greater are a salesperson's perceptions of (a) distributive fairness, (b) procedural fairness, and (c) interactional fairness.

Performance measure appropriateness. Fundamental to a fair merit pay system is the use of credible, appropriate measures of performance (Lawler 1990). If input measures are problematic, the input-outcome ratio will likely be compromised. When employees consider measures inappropriate, the implication is that supervisors either are not evaluating certain job facets that are important for a salesperson's success or are not using measures that capture critical job facets well (Pettijohn, Pettijohn, and Taylor 2000). For example, although customer orientation and satisfaction may be relevant measures of performance, organizations may focus exclusively on sales volume when making pay-raise decisions (Brown and Peterson 1993; Lawler 1990; Mowen et al. 1985). Another negative consequence of inappropriate measures is performance incongruence, whereby employees' evaluations of their performance are not congruent with their supervisor's evaluations (Ramaswami 1996), which results in lower procedural fairness perceptions. Thus:

P₄: The more appropriate the performance measures that are used in merit pay rewards, the greater are a salesperson's perceptions of (a) procedural fairness and (b) distributive fairness.

Participation. Recent studies have provided mounting evidence that the information generated through participatory goal-setting may provide employees with a clearer understanding of their tasks, goals, and expectations (Goodson and McGhee 1991; Sashkin and Williams 1990). As such, employees who participate in the merit pay process improve their attitudes toward the process, making the procedures appear more fair (Cascardi, Poythress, and Hall 2000). Moreover, participation fulfills a salesperson's desire to be heard regardless of whether the expression influences the supervisor. Voice is important to salespeople, because it signals that the supervisor values their input (Kumar, Steenkamp, and Scheer 1995). Using group affiliation arguments, Tyler (1989) suggests that because participation supports employees' beliefs that the supervisor is interested in enhancing their sense of self-respect, participation contributes to greater perceptions of interactional fairness. Thus:

P₅: The greater the participation in decision making, the greater are a salesperson's perceptions of (a) procedural fairness and (b) interactional fairness.

Sample

We obtained data from salespeople employed by a Fortune 500 organization that uses a regional sales structure in which each salesperson reports to an area sales manager. Salespeople sell directly to large customers and use a distributor structure to reach small customers. Although salespeople are involved in account servicing, their primary responsibility is selling. The company requires an annual performance review for input into a merit pay-raise decision. Salespeople are also eligible for sales commission based on their relative performance in their market region. The salary component of the overall compensation is approximately 80%-90%. As such, the magnitude and fairness of annual merit raises is a significant matter. In informal interviews, it was suggested that the merit pay raises range from 3% to 20% of the base salary; however, because of the sensitivity of the information and for reasons of confidentiality, the company refused to allow collection of quantitative pay information through the survey instrument.

At the time of the study, the organization employed 167 salespeople. All salespeople were mailed the survey instrument with a letter that guaranteed confidentiality. Given the study's focus on merit pay decisions, maintaining the confidentiality of the individual responses was a major issue in order to ensure participation and to obtain high-quality data. The respondents were instructed to mail the completed survey in a sealed envelope directly to the researchers.

We obtained responses from 165 salespeople; however, we could not include 11 responses because of missing values, yielding a usable response rate of 92.2%. Because of the relatively small sales force, our pretest discussions with salespeople alerted us to the concern that obtaining customary demographic data (e.g., age, years of experience, gender) might compromise respondent confidentiality. Consequently, in accord with our confidentiality agreement with the sponsoring company, we did not obtain demographic and background data. However, according to company management, the sales force was predominantly male, 30 to 45 years of age, and college educated.

Measurement of Study Variables

For all study constructs, we directly borrowed or adapted the scale items from the literature. In addition, we considered minor wording changes on the basis of comments made by the organization's senior management. The Appendix lists the operational items we used for each construct, and Table 1 provides the univariate statistics for the constructs and the intercorrelations among them.

Fairness dimensions. We measured distributive fairness using a four-item scale that captured the degree to which employees perceive the pay outcomes they received as fair. These items are based on those of Folger and Konovsky (1989) and Greenberg (1986) and pertain to the degree to which employees believe that their pay raise is fair and provides the full amount that they deserved or expected. Procedural fairness pertains to the degree of control or influence afforded to the employee by the process the supervisor uses in arriving at the outcome decision. We measured this concept using a four-item scale that captured both process control, or the degree to which employees have the opportunity to provide input for the decision, and decision control, or the degree to which employees are able to influence the pay decision directly. Interactional fairness items focus on the supervisor's interpersonal behavior. Specific items we used included the degree to which the supervisor was sensitive to employees' needs, considered employees' rights, and dealt with employees in an honest and dignified manner. We drew these items from the work of Folger and Konovsky (1989) and Moorman (1991). A confirmatory factor analysis of fairness items with an a priori hypothesized three-factor model and loading structure produced acceptable fit statistics: χ² = 50.8, degrees of freedom (d.f.) = 74, p > .98, comparative fit index (CFI) = 1.0, normed fit index (NFI) = .97, standardized root mean square residual (SRMR) = .042, and root mean square error of approximation (RMSEA) = 0 (90% confidence interval [CI] of .0 to .028). All estimated loadings were substantively and statistically significant (values > .6 and p < .01, respectively), indicating convergent validity. Consistent with this, we estimated the reliabilities for the distributive, procedural, and interactional fairness dimensions at .89, .78, and .92, respectively. The estimated intercorrelations among the three fairness dimensions range from .50 (distributive-procedural) to .42 (distributive-interactional), indicating that less than 25% variance is shared among them and thus providing initial evidence of discriminant validity. In a recent meta-analysis, Colquitt and colleagues (2001, pp. 437-38) report that the corrected intercorrelations among the fairness dimensions range from .38 to .56; moreover, they note, "our review showed that ... justice [dimensions] have distinct correlates, and measuring the three separately allows for further differences among the dimensions to be examined." Our measures of fairness dimensions yield psychometric evidence that coheres with Colquitt and colleagues' meta-analysis.

Mediators. We measured supervisory trust using a three-item scale developed by O'Reilly and Roberts (1974) and used extensively in previous research. This scale showed a high level of internal consistency, with reliability of .88. We measured job satisfaction using a four-item scale developed by Lucas and colleagues (1987) with some modifications to enhance its contextual relevance. This scale has a reliability value of .94.

Outcomes. We used Mowday, Steers, and Porter's (1979) nine-item organizational commitment scale. We dropped three items because of low loadings. The reliability for the six-item reduced scale is .75. We obtained sales performance evaluations from the supervisor. Because self-reported measures of performance are often subject to biases (e.g., self-presentation), supervisor evaluations are useful. The response rate from supervisors was 100%. Supervisors rated salespeople on three goals set by the organization: sales target performance, business growth, and professional growth. Supervisors also provided overall evaluations of each salesperson. The internal consistency estimate for this performance construct was .93. We measured opportunistic behavior using Jaworski and MacInnis's (1989) six-item scale. The dysfunctional actions captured in this scale include smoothing, focusing, and invalid data reporting. We dropped two items because of low construct loadings. Overall, the four-item measure has an acceptable level of reliability at .75.

Fairness antecedents. We measured linkage to rewards using a three-item scale that assesses whether employees perceive pay raises as directly linked to sales performance and how performance compares to the goals. However, we dropped one of the items, because it did not converge with the other two items. The reliability for the two-item scale was .75. We measured consistent/unbiased application of performance standards using two items that measure the extent to which the merit increases are based on organizational politicking or on the quantity and quality of work performed. We drew these two items from Tyler's (1989) work, and their reliability was .83. We drew the three items that measure performance improvement plan from the work of Folger and Konovsky (1989). These items refer to supervisors' efforts to improve employees' future performance, including discussions of ways to improve performance and use of that information in developing future plans. The alpha reliability for this scale was .70. For a measure to be appropriate, it should track performance for all relevant activities of an employee's job, be based on a thorough analysis of each activity, have a norm or standard for comparison, and be precise. We specifically developed four items to capture these aspects of measure appropriateness. The reliability for this scale was .83. Pay raises are typically determined on the basis of the extent to which an employee meets and exceeds goals. On-the-job goal determination is usually a joint decision between the employee and the supervisor and involves bargaining in a give-and-take environment. Participation refers to the employee's involvement in the goal-setting process. We based these measures on the work of Vroom (1964) and Teas (1981); the reliability for this scale was .83.

Method of Analyses

In testing the proposed hypotheses, we used an analytical method that was sensitive to three issues: ( 1) confounding effects of measurement error, ( 2) potential for misspecification bias, and ( 3) test for mediation effects. In regard to measurement error, we were concerned that the presence of random error would bias the estimation of structural paths unpredictably (Bollen 1989). Although this favors the use of structural equations modeling, we were concerned about its stability and power in light of our sample size (N = 154). To strike a balance between these concerns, we used an approach based on Bagozzi and Edwards's (1998) suggestion of partial disaggregated models. Specifically, we used two composites formed by combining the odd-and even-numbered items wherever possible as indicators for each latent construct. Bagozzi and Edwards show that partial disaggregated models are likely to have better statistical properties than approaches that use individual measures as indicators. Such models perform reasonably well, compared with fully disaggregated models, in terms of control over measurement error. As in structural equations models, the estimated coefficients reflect relationships among underlying theoretical constructs and are "adjusted" for measurement error. In addition, they provide a systematic basis for evaluating the "fit" of the hypothesized model to data (e.g., χ² statistic, incremental fit indexes, RMSEA; Bentler 1995).

Misspecification bias can occur if some of the unhypothesized effects are significant and not included in the empirical analysis. Specifically, the proposed theoretical model includes a system of effects wherein the fairness antecedents (Level I) influence fairness evaluations (Level II), which in turn affect posited mediators (Level III), and finally the mediators influence outcomes (Level IV). Hypotheses that link Level I and Level III, for example, are not included. To examine the significance of such nonhypothesized direct effects systematically without forsaking parsimony, we used the proposed model as the baseline model and tested for the significance of incremental increases in model fit due to nonhypothesized direct effects. This involved computing a change in the χ² statistic and the corresponding change in degrees of freedom. To streamline this procedure, we used sets of effects that represented different levels in the model. For example, we included direct effects from all fairness antecedents (Level I) to job satisfaction and supervisor trust (Level III) and the change in χ² tested with 10 degrees of freedom. We examined individual coefficients to retain the significant effects for the next step of analysis. We systematically implemented this procedure to test all potential unhypothesized effects and mitigate misspecification bias.

Finally, we were concerned about providing a test for the mediation hypotheses. Our model hypothesizes that supervisor trust and job satisfaction mediate the effect of fairness evaluations on job outcomes. To test mediation effects, we followed the procedures Baron and Kenny (1986) suggest. Specifically, we estimated a "direct" model in which we eliminated mediation variables and estimated direct effects. We then compared the direct effects with the corresponding coefficients from a model that included the mediating variables. A full mediation was indicated if ( 1) the "direct" effects model produced a significant effect on a given outcome, ( 2) the corresponding direct effect was reduced to insignificance after inclusion of the mediating variable, and ( 3) the mediator had a significant effect on the focal outcome. Mediation was not indicated when the direct effect remains virtually unchanged in Step 2. Finally, partial mediation was indicated when the direct effect in Step 2 is reduced but does not become nonsignificant. In addition, given the sample size of 154 and a complex model that involved interrelationships among 13 distinct constructs, we were concerned about the power of statistical tests at the customary level of significance (.05). Consequently, we use a 10% level of significance for statistical testing.

Measurement Model Analysis

Before testing the hypothesized model, we estimated a fully disaggregated measurement model with all observed indicators to ensure that the measures corresponded only to their hypothesized constructs and evidenced acceptable reliability as well as convergent and discriminant validity. Using the confirmatory factor analysis procedures available in EQS, we estimated a measurement model that included all 51 items that we hypothesized to measure the 13 study constructs. We proposed the individual measures to load only on a single factor, in accord with conceptual definitions. This measurement model produced the following fit statistics: χ² = 1307, d.f. = 1147, p < .01, NFI = .90, NNFI = .99, CFI = .99, SRMR = .062, and RMSEA (90% CI) = .032 (.021 to .039). Although the χ² statistic is significant, the other indicators of relative and absolute fit (e.g., NFI, CFI, RMSEA) and the indicator for parsimonious fit (e.g., NNFI) unequivocally suggest that the hypothesized measurement model is a reasonably good representation of the variance-covariance matrix of study measures. The estimated parameter estimates in Table 2 reveal that the standardized factor loadings, without exception, are statistically significant (t-values > 2, p < .05) and substantively large (>.30). In addition, the composite reliability estimates exceed .70, and variance extracted exceeds .50, with two exceptions that involve improvement plan and commitment. Note that the composite reliability estimates differ slightly from alpha reliability estimates we provided previously; the former are based on maximum likelihood estimates. Finally, the average factor correlation is .33, indicating that, on average, less than 11% of the variance is shared among the constructs. Overall, our results suggest that the modeled constructs have reasonable psychometric properties and appear suitable for substantive analysis and interpretation.( n1)

Overall Fit of the Hypothesized Structural Model

Initially, we estimated the hypothesized model of Figure 1. We encountered no particular problems in estimation and achieved convergence without any boundary conditions. The estimated coefficients from this model are listed in Table 3 as "initial" coefficients. In accord with our analytical plan, we then examined the potential for misspecification bias by estimating three less restrictive models in sequential steps that systematically allowed for direct effects between nonadjacent levels of variables. After each step, we retained significant coefficients for the next step of analysis. In all, we included one path in Step 1 that involved measure appropriateness (effect on interactional fairness), four paths in Step 2 that involved distributive fairness (effect on performance and opportunistic behaviors) and participation (significant effect on satisfaction and trust), and one path in Step 3 (improvement plan → performance). The final model yielded fit statistics as follows: χ² = 276.7, d.f. = 258, p = .20, NFI = .94, NNFI = .99, CFI = .99, SRMR = .051, RMSEA (90% CI) = .023 (.001 to .040), and Akaike information criterion (AIC) = -239.3 (independence AIC = 3699.2). Given the nonsignificant χ² goodness-of-fit test and other fit indicators, it appears that the final model provides an acceptable representation of the data.

Structural Coefficients and Hypotheses Tests

Table 3 provides the estimated coefficients from the final model, and Figure 2 displays these results graphically. Overall, it appears that the final model provides a reasonable explanation for distributive and interaction fairness (R²= .48 and .58, respectively) but not procedural fairness (R² = .12). Likewise, we achieved meaningful explanation levels for performance and commitment (R² = .40 and .42, respectively) but not opportunistic behaviors (R² = .17). Finally, the explanation levels for supervisor trust and job satisfaction are also reasonable (R² = .69 and .38, respectively).

Regarding H₁, Table 3 reveals that interactional fairness has a strong positive effect on supervisor trust (β = .46, p < .01); however, the effect of procedural fairness is nonsignificant. Moreover, contrary to H₂, distributive fairness is significantly associated with supervisor trust (β = .18, p < .01). Although not posited, participation is positively associated with supervisor trust (β = .34, p < .01). Moreover, as we note in Table 3, the initial coefficient for interactional fairness is significantly higher, indicating that its effect is overestimated if the model is not respecified to account for participation.

Regarding job satisfaction, we found interactional fairness to enhance a salesperson's job satisfaction significantly, in accord with H₃ (β = .19, p < .05). Neither distributive nor procedural fairness yielded a significant effect; however, participation produced a significant influence on job satisfaction (β = .54, p < .01). As in the case of supervisor trust, the initial coefficient for interactional fairness is more than twice as great (β = .54 versus .19), underscoring the potential misspecification as a result of the direct effects of participation on job satisfaction.

In partial support of H₄, supervisor trust significantly enhances the commitment of salespeople (β = .16, p < .05) and diminishes their tendency to engage in opportunistic behaviors (β = -.52, p < .01). However, the performance of salespeople is unaffected by supervisor trust. Notably, the initial estimated coefficients indicate that the influence of supervisor trust is significant (β = .50); however, this effect reduces to nonsignificance (β = .13) when the model is respecified to account for the direct effects of distributive fairness and performance improvement plans. Consistent with H5, salespeople's job satisfaction positively affects their commitment (β = .57, p < .01), but it fails to curb their opportunistically oriented behaviors. In addition, in accord with H6, we found that job satisfaction does not have a significant affect on performance; however, distributive fairness and improvement plan significantly and positively influence salesperson's performance (β = .28 and .35, p < .01). In addition, distributive fairness has a positive, significant effect on opportunistic behaviors (β = .29, p < .01).

Finally, to test for mediation, we estimated a direct effects model that excluded the hypothesized mediating variables of supervisor trust and job satisfaction. The model yielded the following fit statistics: χ² = 401.2, d.f. = 265, p < .001, NFI = .91, NNFI = .96, CFI = .97, SRMR = .17, RMSEA (90% CI) = .06 (.047 to .07), and AIC = -128.6. Compared with the mediated model, the direct effects model indicates a significant and substantive deterioration in model fit (χ²_difference = 125.2, d.f. = 7, p < .001), indicating that the proposed mediators play a significant role in fairness mechanisms. In addition, for organizational commitment, a significant direct effect emerges for interactional fairness (β = .34, p < .01). Because this direct effect becomes nonsignificant for the mediated model and other conditions for mediation are met (significance of interactional fairness → job satisfaction, and job satisfaction → commitment paths; see Table 3), we can conclude that the effect of fairness judgments on commitment is fully mediated. In contrast, for performance, significant direct effects emerge for distributive fairness and improvement plan (β = .31 and .38, respectively, p < .01) that remain significant in the mediated model (see Table 3, corresponding values of .28 and .35). Furthermore, because none of the mediation conditions is satisfied, the effects of fairness judgments on job performance are unmediated. We obtained partial mediation for opportunistic behaviors, because though the direct effects of interactional and distributive fairness are significant (β = -.20 and .16, respectively, p < .05), only the latter remains significant in the mediated model. Combined with the significant effects obtained in the mediated model for the interactional fairness → trust and the trust → opportunistic behavior paths, we can conclude that the effect of interactional fairness on opportunistic behaviors is fully mediated, but the effect of distributive fairness is not. Overall, the effect of fairness judgments on opportunistic behaviors is partially mediated.

In terms of propositions, Table 3 indicates that the antecedents significantly influence distributive fairness. Consistent with P1, the linkage to rewards is positively and significantly associated with distributive fairness (β = .25, p < .01). Likewise, the distributive fairness perceived by salespeople is significantly enhanced when they view the merit pay decision-making procedures as unbiased and when supervisors work with them to develop plans for performance improvement (β = .33 and .20, respectively, p < .05). These results support P₂ and P₃. Consistent with P₄, measure appropriateness is positively associated with distributive fairness (β = .18, p < .10).

In contrast, the included antecedents are less effective in accounting for the variability in procedural fairness. Of the hypothesized antecedents, the use of consistent, unbiased decision-making procedures and participation have a significant influence on procedural fairness (β = .19 and .18, respectively, p < .10). These results support P2 and P5 but not P₃ or P₄.

For interactional fairness, when supervisors help develop performance improvement plans, salespeople's perceptions of interactional fairness are enhanced (β = .34, p < .01). Participation in decision making is also associated with greater interactional fairness (β = .32, p < .01). These results support P₃ and P₅. Finally, measure appropriateness enhances interactional fairness (β = .25, p < .01).

This study was motivated by three objectives: ( 1) to use a broader, more complete conceptualization of the merit pay fairness construct that included dimensions of distributive, procedural, and interactional fairness; ( 2) to study the simultaneous mediating mechanisms of supervisory trust and job satisfaction in fairness processes; and ( 3) to explore the supervisory behaviors during the merit pay decision process that increase fairness perceptions. Our findings offer initial insights into these issues and provide concrete directions for further research and managerial guidelines. Before discussing these findings, we address the limitations of our study.

Limitations

Several limitations of our study are noteworthy. First, the study is based on cross-sectional survey data, and we advise caution in drawing cause-effect inferences. In addition, the association among constructs may be inflated as a result of common method variance. To reduce this inflation, we followed Heneman's (1974) suggestions for enhancing the accuracy and validity of self-ratings by guaranteeing confidentiality. Furthermore, we collected data for salespeople's performance from a different source (i.e., supervisors). The use of multiple source data likely reduces the influence of common method bias. Nevertheless, because we focus on the differential effects of constructs (e.g., interactional and distributive fairness), we recognize that the common method variance would act to obscure such differential effects. In this sense, our findings of differential patterns are likely conservative. Second, the study is based on a sample of salespeople employed by a single organization. Studies in other contexts, such as across organizations and over time, are needed to establish the generalizability of our findings. Third, two of the study constructs (commitment and improvement plan) have less-than-ideal psychometric properties, with several low-loading items. Low loadings result in low reliability that introduces random noise in the results, thereby making it difficult to detect significant effects. Nevertheless, the confirmatory and exploratory factor analysis results confirm that the measures are not confounded. Fourth, the actual merit pay-raise data for the sample were not available. It is possible that salespeople who receive higher-than-average raises evidence a positive bias. However, we were able to test for this bias indirectly by using perceived favorability of outcomes as a covariate in the analysis; we found that our results were virtually unaltered.

Despite these limitations, our results offer useful insights into the differential effects of fairness dimensions by considering the simultaneous influences of all three fairness judgments, direct and mediated pathways to satisfaction and supervisor trust that link fairness judgments and outcomes, and potential relationships among antecedents and fairness judgments. We discuss each of these contributions. Thereafter, we draw managerial implications.

Differential Effects of Fairness Dimensions

The results of our study reveal a differential pattern of effects for the three fairness dimensions. Although interactional fairness significantly influences supervisor trust and job satisfaction, distributive fairness has a significant effect on supervisor trust but not on job satisfaction. It appears that when the effect of interaction quality between salespeople and their supervisors is accounted for, presence or absence of distributive fairness is unimportant for generating job satisfaction.( n2) This result is counter to the findings of a study by Netemeyer and colleagues (1997). However, because Netemeyer and colleagues did not consider interactional fairness, it is difficult to predict whether the impact of distributive fairness that they observed could potentially be weakened with a complete accounting of fairness dimensions.

More important, procedural fairness is associated with neither supervisor trust nor job satisfaction. This nonsignificance of procedural fairness in our study is at variance with results from previous studies (Konovsky and Pugh 1994; Kumar, Scheer, and Steenkamp 1995). However, most previous studies have used a composite view of procedural fairness, including decision-making structures and supervisors' enactment of these structures (Aryee, Budhwar, and Chen 2002). Our findings suggest that when procedures and their enactment are separated, it is interactional fairness that affects work outcomes. The inference that can be drawn from this finding is that the effects observed for procedural fairness in (some) previous studies may be attributable to the interactional fairness component: Why should interactional fairness be more important than procedural fairness? It is possible that salespeople have reasonably complete information about how the supervisor interacted with them (because of the information's relative transparency), but they do not have complete information on the quality of procedures used by the supervisor to arrive at merit pay decisions (because of its relative confidentiality). In addition, interactional fairness may be more potent not only because of its intrinsic value (e.g., treating salespeople with dignity) but also because of its signaling value (e.g., as a "signal" for the procedures used). The validity of these conjectural explanations needs to be established by further replication and extension of our work, yet we appear to have sufficient evidence to suggest that studies that fail to include all three fairness dimensions risk misspecification bias.

Mediated and Direct Pathways of Fairness: Outcomes Relationships

Although we posited that the influence of fairness judgments on salesperson outcomes is mediated by supervisor trust and job satisfaction, our results offer a complex pattern of evidence. Overall, fairness effects on commitment are indirect and fully mediated, whereas opportunistic behaviors are influenced by partially mediated pathways that involve a combination of direct and indirect effects. In contrast, fairness effects on performance are direct and unmediated. We discuss each in turn.

In terms of salespeople's commitment to their organizations, the direct influence of fairness perceptions is marginal; rather, commitment is driven primarily by salespeople's job satisfaction and, to a lesser degree, by supervisor trust.( n3) This suggests that the way employees feel about their job experiences is more important to organizational loyalty than the level of trust they have with their supervisors. Because satisfaction and trust, in turn, are mostly influenced by interactional fairness, it appears that organizational loyalty is influenced more by whether an organization treats its employees with respect and dignity and less by whether it provides fair distribution of monetary rewards.

A salesperson's propensity to engage in opportunistic behaviors is influenced directly by supervisor trust and distributive fairness.( n4) When trust is established through dignified and respectful treatment of the employee (e.g., interactional fairness and participation), salespeople may view the supervisor as a partner rather than an adversary and may work actively to make their job less difficult. It appears that employees do this by avoiding opportunistic behaviors that may otherwise demand valuable supervisory time for close monitoring of employee activities (Bateman and Organ 1983). As a result, the sales manager is left to perform strategic functions rather than become distracted in time-consuming, nonproductive monitoring behaviors. The results indicate a direct, positive effect of distributive fairness, suggesting that higher fairness of merit pay awards increases the propensity of opportunistic behaviors. This is counterintuitive. However, distributive fairness has an indirect, negative effect on opportunistic behaviors through its effect on supervisor trust. The direct and indirect effects are in opposition such that the net effect of distributive fairness is a weak, marginally positive effect (.15). A possible explanation for this positive net effect is that salespeople may not be satisfied with the absolute merit pay distributions, though they remain satisfied with the relative distributions. In turn, this dissatisfaction may prompt salespeople to engage in self-interest-seeking opportunistic behaviors. Because of restrictions on the measurement process imposed by the organization, it was not feasible to collect information on the actual pay-raise amount, and thus we cannot evaluate this explanation in the present setting. Further research needs to examine fairness of both relative and absolute distributions and their relative effects.

Finally, in terms of job performance, although neither satisfaction nor trust has a significant, direct influence, both distributive fairness and performance improvement plans yield direct, unmediated effects. The lack of a significant satisfaction-performance association is consistent with previous research; however, the lack of trust effect is somewhat surprising. There are two possible explanations for this result. The first, and probably less viable, explanation is that salesperson performance is under the direct regulation of both merit pay awards (i.e., distributive fairness) and performance enhancement plans and is less sensitive to social exchange mechanisms. The second, and probably more viable, explanation is that when the supervisor acts in an equitable manner and provides employees with feedback that can improve their future performance, trust may have a lesser role in encouraging reciprocal actions. In this sense, the social exchange variable, trust, has a significant effect, but only in a fairness deficient environment. Further research is needed to test these speculations by examining potential moderators of the trust → performance relationship.

In addition, our study sheds new light on an outstanding issue in the literature related to the (lack of) relationship between salesperson job performance and satisfaction. Our results reveal that though satisfaction and performance are weakly correlated, the explanation levels are relatively high for both at 38% and 40%, respectively. Participation and interactional fairness have a strong effect on job satisfaction, and performance improvement plan and distributive fairness have a strong effect on job performance. As such, it appears that satisfaction and performance are linked to different fairness mechanisms. Salespeople's job satisfaction is sensitive to the aspects of their relationship with the supervisor and work that "touches their heart," such as when the supervisor engages in behaviors that signal respect, dignity, and participatory decision making. In contrast, salesperson performance is apparently more sensitive to "head" factors, including the supervisor's attempts to aid in developing a plan for individual performance improvement and the hard calculus of distributive fairness. The notion that job satisfaction and performance may be related to "heart" and "head" mechanisms that are rooted in different aspects of fairness perceptions offers new insight and directions for further research.

Antecedents of Fairness Perceptions

Our objective was to use this study's empirical results to provide a theoretical foundation for the study of fairness antecedents. Using theoretical and pragmatic considerations, we included five antecedents for our initial examination. Our results demonstrate that the included antecedent factors are meaningful in understanding the formulation of distributive and interactional fairness judgments in merit pay decisions (R² = 48% and 58%, respectively). However, in regard to procedural fairness, our results indicate a more tentative stance (R² = 12%). We discuss each result in turn.

Distributive fairness. Results show that distributive fairness perceptions are greater when ( 1) the merit pay rewards that salespeople receive are linked to job performance, ( 2) the supervisor is unbiased and consistent in applying appropriate performance standards, and ( 3) the supervisor aids in designing a plan for improving a salesperson's future performance. Taken together, these results imply that after the impact of the linkage between current input and outcomes is accounted for, the prospect of improved future outcomes influences salespeople's beliefs about the fairness of their current reward decision. This indicates that unfairness perceptions arising from current distributions could be mitigated if supervisors were to begin discussing performance details post hoc and outlining how performance can be improved in the future. As such, our results contradict conventional wisdom that salespeople are concerned solely about short-term rewards. Evidently, even though salespeople respond to current merit rewards, they are responsive to strategies for enhancing future rewards.

Procedural fairness. Two of the four proposed antecedents for procedural fairness (use of unbiased procedures and participation) were statistically significant. Neither use of appropriate data nor supervisory focus on future performance was significant. Consequently, it appears critical that sales managers engage in behaviors and practices that emphasize an unbiased approach in merit pay decisions. An option is to produce and share hard evidence that supports that performance-pay linkages are not compromised by individual manager biases. In cases in which performance can be reasonably quantified, it may be useful to compute and monitor performance-pay correlation with the notion that attenuation of this correlation beyond a cutoff value (e.g., .80) is probably on account of process errors. Although contextual factors must be considered in designing such strategies, our findings argue for management behaviors and practices that increase transparency of merit pay decisions and promote neutrality by curbing individual biases. The significance of participation rests on it being a means of influencing the merit pay decision-making process. Employees who participate may believe that they were able to manage their supervisor's expectations when making pay decisions. It is inevitable that employees will sometimes receive unfavorable outcomes, and participation can temper the potential backlash from employees when this occurs.

The nonfindings for performance measure appropriateness and improvement plan are intriguing. Use of appropriate measures enables supervisors to focus on job activities that are important. Apparently, although they may use appropriate measures, supervisors can still show bias in translating individual measurements into rewards. Likewise, it was believed that when supervisors discuss improvement plans with employees, the latter would be presented with the opportunity to provide input and thereby influence the performance evaluation process. Then, why is it that the design of improvement plans would influence distributive fairness but not procedural fairness? Is it possible that salespeople consider these plans instrumental in achieving greater performance but not necessarily in increasing their influence over the reward decision? More research is needed to unravel these intriguing effects.

Interactional fairness. Three antecedents (improvement plans, measure appropriateness, and participation) are strongly related to interactional fairness. Developing performance improvement plans not only helps in aligning perceptions and expectations between supervisors and employees (Morgan and Hunt 1994) but also signals that the supervisor cares. Using appropriate measures indicates that the supervisor understands that the salesperson has the right to be evaluated correctly. Finally, participation reinforces the social standing of salespeople in their respective groups and informs that they are valued members of the sales team.

Overall, the supported propositions in regard to fairness antecedents suggest that these linkages are meaningful, and despite the limited number of antecedents that we could include in our study, they together explain a significant proportion of the variance in distributive and interactional fairness. Although the explanation level for procedural fairness is more modest, the pattern of results appears reasonable and offers a fertile ground for future theorizing.

Managerial Implications

A major finding of potential interest to managers is the dominant role of interactional fairness in achieving two of the three outcomes that govern organizational effectiveness. It was suggested previously that supervisory interactions not only might be more transparent than procedures but also might signal to employees that the supervisor cares for their well-being. This is good news for managers, because the economic costs of interacting in a manner that raises the dignity and standing of employees are not likely to be as high as the costs associated with satisfying either procedural or distributive fairness. This does not imply that the distribution of rewards or the procedures used to determine them are not important in their own right. Our point is merely that interactional fairness is critical in fairness mechanisms and one that can be easily achieved by managers.

Another major finding of the study is the importance of distributive fairness in influencing employees' job performance. The sense of equity that arises from being rewarded fairly provides the incentive to salespeople to work harder and to improve their job performance. In a way, this result argues for unveiling the shroud of secrecy that typically exists in pay-raise situations. Previous research has noted that when employees are not provided comparative pay information, they tend to believe that they obtained a raise that is lower than others, which in turn typically results in lower fairness perceptions (Futrell and Jenkins 1978).

From a managerial perspective, it would be especially unfortunate to interpret our results to imply that procedural fairness may be safely ignored: The results are more complex. Often, in the analysis of the sort used herein, the relative importance of a variable is determined by the principle of discrimination, that is, the variable along which the respondents can be differentiated the most. As such, the appropriate interpretation is that given the situational context of our sample, further increases in procedural fairness may be less potent than similar increases in interactional fairness. Furthermore, varying interactional fairness holds more promise given the current situation and pay practices for the specific sample of salespeople used. Managers are cautioned to develop such understanding in the light of their specific context and sales force practices.

What can managers do to shape employee perceptions of the three fairness dimensions? They can put into place a performance improvement plan for employees. Such plans are opportunities for supervisors to provide feedback to salespeople about what they are doing "right" or "wrong" and to explicate what they should do to improve performance. The positive effect of such planning suggests that salespeople do not view it as a performance impediment resulting from supervisors stepping on their turf. A clear message from the results is that managers need to spend more time with salespeople with an aim to improve their future performance ability.

Managers could also allow salespeople to participate in goal setting and process determination, because this has a favorable influence on not only interactional and procedural fairness but also the two mediators. Participation is the backbone behind the concept of "quality circles" in total quality management. As our results suggest, it may also be the backbone behind the "loyalty" of salespeople to their supervisors and organizations. It appears that employees expect to be asked to participate (based on influence on procedural fairness) and derive great value from the symbolism this represents (based on its influence on interactional fairness). Employees are being asked to do more with less, and participation may be a way by which employees can be made to believe that they are being given more. Finally, managers need to ensure that pay is linked to performance and to make that linkage known to employees. They may be able to do this not only by using appropriate measures of performance but also by ensuring that no employee has undue influence over the rewards process.

The study of fairness judgments and its consequences in sales force settings dates to the mid-1980s, and a considerable body of work exists that establishes the relevance and significance of such judgments. Our study advances this body of work by posing different questions and shifting the direction of inquiry. Instead of asking whether fairness judgments matter for sales force outcomes, we ask, How do fairness judgments work to influence critical outcomes and what factors influence the formation of fairness judgments? This shift has important theoretical and managerial implications.

Theoretically, by highlighting the contrasting mechanisms by which distributive, procedural, and interactional fairness judgments influence job outcomes, our findings implore future researchers to eschew studies that fail to capture the breadth of the fairness construct and/or complexity of its mechanisms. Significantly, our work opens new dialogues on several unresolved issues in sales force management. We presented arguments about why performance and job satisfaction, two seemingly related concepts, may be driven by different fairness mechanisms. Likewise, although our study clarifies that pay-to-performance linkages work, it also highlights the performance-commitment dilemma; that is, although distributive fairness and appropriate pay-for-performance linkages may propel salespeople to perform better, they do little to gain their commitment. Organizational commitment is influenced by job satisfaction; in our data, job satisfaction is shaped mainly by interactional fairness. Consequently, to motivate salespeople toward higher performance and retain them requires that two separate mechanisms be activated simultaneously, one involving interactional fairness and job satisfaction and the other involving distributive fairness and improvement plans. Activation of just one mechanism produces obvious pitfalls such as losing the high performers or retaining the low and average performers. This requires managers to focus simultaneously on multiple dimensions of fairness and activate multiple mechanisms. Human agency and judgments are complex, dynamic phenomena that defy simple answers, yet this complexity and dynamism can be understood and captured with the tools at the disposal of marketing science today. Our study has provided an initial attempt in the context of merit pay-raise decisions. Future researchers and managers will find our proposed fairness model a fertile ground for further refinement and development to understand how to motivate and retain a high-performance sales force.

The author names are listed in alphabetical order; both authors contributed equally to the research effort.

The authors thank Stephen A. Gorton, an independent consultant, for his help with data collection.

(n1) Based on the criterion of comparing average variance extracted with the highest variance shared, two subsets of constructs involving (1) commitment, satisfaction, and improvement plan and (2) participation, interactional fairness, and trust failed to provide clear evidence of discriminant validity. To check this further, we performed exploratory factor analyses of these subsets to ensure that the measures are not confounded and that interfactor correlations are far from unity. In each case, we obtained supporting and unequivocal evidence. Specifically, the items have a dominant loading on the hypothesized factor, and cross-loadings, when present, are relatively smaller in magnitude. Moreover, interfactor correlations are all less than .65.

(n2) To further test for this differential effect, we tested a constrained model in which the three fairness dimensions were constrained to have equal effect on job satisfaction. This analysis revealed that though the invariance hypothesis cannot be rejected for distributive and procedural fairness (χ² = .47, p = .7), the influence of interactional fairness is indeed distinct (χ² = 3.80, p = .05). Likewise, when the effects of procedural and interactional fairness on supervisor trust were constrained to be equal, we obtained a significant χ² indicating rejection of this hypothesis (χ² = 13.1, p < .01).

(n3) To test for this differential effect, we constrained the effects of supervisor trust and satisfaction on commitment to be equal. This yielded a significant statistic, indicating a rejection of this proposal (χ² = 4.74, p < .05).

(n4) Likewise, a differential effects test produced a significant statistic, suggesting that opportunistic behaviors are differently related to satisfaction, trust, and distributive fairness (χ² = 10.25, p < .01).

Legend for Chart:

B - PF
C - IF
D - DF
E - ST
F - JS
G - JP
H - OC
I - OB
J - RL
K - CA
L - IP
M - MA
N - PT
O - Mean
P - Standard Deviation

            A                  B       C       D          E
                               F       G       H          I
                               J       K       L          M
                                       N       O          P

Procedural fairness(PF)       1.00
                                              2.94(a)     .88

Interactional fairness(IF)     .46    1.00
                                              4.23(b)     .75

Distributive fairness(DF)      .50     .42    1.00
                                              3.01(b)    1.06

Trust(ST)                      .48     .74     .44       1.00
                                              5.29(c)    1.45

Job satisfaction(JS)           .19     .38     .11        .34
                              1.00
                                              4.47(b)     .59

Job performance(JP)            .18     .29     .27        .28
                               .04    1.00
                                              3.57(b)     .98

Organizational
commitment(OC)                 .14     .29     .16        .28
                               .54     .08    1.00
                                              4.48(b)     .44

Opportunistic behavior(OB)     .07    -.13     .03       -.23
                              -.04    -.03     .02       1.00
                                              2.83(b)     .83

Linkage to rewards(RL)         .36     .36     .42        .35
                               .23     .12     .19       -.11
                              1.00
                                              3.53(b)    1.02

Consistent/ unbiased
application(CA)                .08     .10     .04       -.04
                               .14    -.05     .14       -.14
                               .56    1.00
                                              2.84(d)     .71

Improvement plan(IP)           .58     .54     .44        .39
                               .27     .02     .24        .04
                               .39     .16    1.00
                                              3.82(b)     .84

Appropriateness of
measures(MA)                   .45     .54     .44        .45
                               .29     .20     .22        .04
                               .36     .05     .54       1.00
                                              3.61(b)     .82

Participation(PT)              .44     .52     .28        .55
                               .51     .24     .31       -.17
                               .25     .08     .37        .37
                                      1.00    3.80(b)     .78

(a) Four-point Likert scale.

(b) Five-point Likert scale.

(c) Seven-point Likert scale.

(d) Three-point scale.

Notes: Correlations less than .15 are nonsignificant (p > .05).

Legend for Chart:

A - Construct/Item
B - Loading
C - t-Value
D - Variance Extracted
E - Highest R²
F - Average R²
G - Reliability

          A              B       C       D      E      F      G

Job Performance

JP1                     .81    11.52    .74    .15    .07    .93

JP2                     .76    10.52

JP3                     .82    11.84

JP4                     .93    14.51

JP5                     .89    13.42

Opportunistic Behaviors

OB1                     .40     4.55    .50    .08    .02    .75

OB2                     .62     7.49

OB3                     .87    11.09

OB4                     .71     8.75

Organizational Commitment

OC1                     .40     4.61    .45    .40    .11    .75

OC2                     .54     6.44

OC3                     .38     4.23

OC4                     .74     9.48

OC5                     .78    10.16

OC6                     .59     7.09

Supervisor Trust

ST1                     .76    10.54    .72    .65    .32    .88

ST2                     .83    12.06

ST3                     .94    14.54

Job Satisfaction

JS1                     .85    12.63    .78    .40    .14    .94

JS2                     .91    14.02

JS3                     .92    14.19

JS4                     .86    12.76

Distributive Fairness

DF1                     .93    14.44    .66    .28    .18    .89

DF2                     .92    14.12

DF3                     .74    10.10

DF4                     .64     8.34

Interactional Fairness

IF1                     .83    12.04    .68    .65    .29    .92

IF2                     .74    10.08

IF3                     .87    12.99

IF4                     .85    12.39

IF5                     .80    11.35

IF6                     .81    11.51

Procedural Fairness

PF1                     .65     8.08    .58    .36    .17    .78

PF2                     .68     8.58

PF3                     .57     6.87

PF4                     .82    10.91

Linkage to Rewards

RL1                     .52     6.65    .62    .24    .13    .75

RL2                     .98    16.30

Consistent/Unbiased Application

CA1                     .91    11.43    .72    .26    .12    .83

CA2                     .78     9.64

Improvement Plan

IP1                     .62     7.46    .44    .47    .24    .70

IP2                     .72     8.93

IP3                     .64     7.73

Measure Appropriateness

MA1                     .77    10.34    .56    .42    .21    .83

MA2                     .83    11.49

MA3                     .73     9.56

MA4                     .66     8.34

Participation

PT1                     .82    11.34    .55    .50    .33    .83

PT2                     .66     8.39

PT3                     .68     8.76

PT4                     .81    11.14

Notes: Loading = standardized coefficient estimate by the
elliptical reweighted least squares method using EQS software.
The t-values greater than 1.645 indicate significant effects at
p = .05 for a one-tailed test. Variance extracted is based on
Fornell and Larcker's (1981) formula. Highest R² is the
highest variance shared between this construct and any other
construct in the model; it is computed as the square of highest
R (correlation). Average R² is the average variance shared
between this construct and all other constructs; it is computed
as the mean of squared correlations. Composite reliability is
based on Fornell and Larcker's formula.

Legend for Chart:

A - Dependent Variable
B - Initial Coefficient
C - Final Coefficient
D - t-Value
E - R²

        A
                        B           C           D         E

Distributive Fairness (DF)
                                                        .48(x)

Linkage to rewards (RL) → DF
                     .27(x)      .25(x)      2.9(x)

Consistent/unbiased application (CA) → DF
                     .31(x)      .33(x)      4.0(x)

Measure appropriateness (MA) → DF
                     .14         .18         1.6

Improvement plan (IP) → DF
                     .23(x)      .20(x)      2.1(x)

Procedural Fairness (PF)
                                                        .12(x)

Consistent/unbiased application (CA) → PF
                     .20(x)      .19(x)      2.1(x)

Improvement plan (IP) → PF
                     .03         .06         .51

Measure appropriateness (MA) → PF
                     .01         .03         .04

Participation (PT) → PF
                     .22(x)      .19         1.6

Interactional Fairness (IF)
                                                        .58(x)

Improvement plan (IP) → IF
                     .45(x)      .34(x)      3.7(x)

Measure appropriateness (MA) → IF
                     --          .25(x)      2.7(x)

Participation (PT) → IF
                     .43(x)      .32(x)      3.4(x)

Supervisor Trust (ST)
                                                        .69(x)

Distributive fairness (DF) → ST
                     --          .18(x)      2.4(x)

Procedural fairness (PF) → ST
                     .06         .02         .10

Interactional fairness (IF) → ST
                     .77(x)      .46(x)      5.1(x)

Participation (PT) → ST
                     --          .34(x)      3.7(x)

Job Satisfaction (JS)
                                                        .38(x)

Distributive fairness (DF) → JS
                     -.11        -.14        -1.4

Procedural fairness (PF) → JS
                     .06         -.09        -1.1

Interactional fairness (IF) → JS
                     .54(x)      .19(x)      1.7(x)

Participation (PT) → JS
                     --          .54(x)      4.7(x)

Job Performance (JP)
                                                        .40(x)

Supervisor trust (ST) → JP
                     .50(x)      .13         1.1

Job satisfaction (JS) → JP
                     --         -.03        -.40

Distributive fairness (DF) → JP
                     --          .28(x)      2.6(x)

Improvement plan (IP) → JP
                     --          .35(x)      3.0(x)

Organizational Commitment (OC)
                                                        .42(x)

Supervisor trust (ST) → OC
                     .16(x)      .16(x)      1.7(x)

Job satisfaction (JS) → OC
                     .58(x)      .57(x)      4.4(x)

Opportunistic Behaviors (OB)
                                                        .17(x)

Supervisor trust (ST) → OB
                    -.33(x)     -.52(x)     -3.2(x)

Job satisfaction (JS) → OB
                     .09         .11         1.1

Distributive fairness (DF) → OB
                     --          .29(x)      2.2(x)

Notes: Initial coefficient is the estimated standardized
coefficient before we respecified the model to account for
misspecification bias. Paths not hypothesized and not tested
for respecification are indicated by a dash. Final coefficient
is the estimated standardized coefficient after we respecified
the model and accounted for misspecification bias. The
corresponding t-values are in the adjacent column; t-values
greater than 1.645 indicate significant effects at p = .05
for a one-tailed test. Coefficients in (x) bold are significant
at p = .05, and coefficients in italics are significant
at p = .10.

DIAGRAM: FIGURE 1 The Conceptual Model for Understanding Fairness Mechanisms and Consequences

DIAGRAM: FIGURE 2 The Estimated Model for Understanding Fairness Mechanisms and Consequences

Anderson, James C. and James Narus (1990), "A Model of Distributor Firm and Manufacturer Firm Working Partnerships," Journal of Marketing, 54 (January), 42-58.

Aryee, Samuel, Pawan S. Budhwar, and Zhen Xiong Chen (2002), "Trust as a Mediator of the Relationship Between Organizational Justice and Work Outcomes: Test of a Social Exchange Model," Journal of Organizational Behavior, 23 (March), 267-85.

Bagozzi, Richard A. and Jeffrey Edwards (1998), "A Generalized Approach for Representing Constructs in Organizational Research," Organizational Research Methods, 1 (January), 45-87.

Baron, Reuben M. and David A. Kenny (1986), "The Moderator-Mediator Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations," Journal of Personality and Social Psychology, 51 (December), 1173-82.

Bartol, Kathryn M. (1999), "Reframing Salesforce Compensation Systems: An Agency Theory-Based Performance Management Perspective," Journal of Personal Selling & Sales Management, 19 (Summer), 1-16.

----- and David C. Martin (1989), "Influence of Managerial Allocation: A Dependency Perspective," Personnel Psychology, 41 (Summer), 361-78.

Bateman, Thomas S. and Dennis W. Organ (1983), "Job Satisfaction and the Good Soldier: The Relationship Between Affect and Employee Citizenship," Academy of Management Journal, 26 (4), 587-95.

Bentler, Peter (1995), EQS: Structural Equations Program Manual. Los Angeles: Multivariate Software.

Bollen, Kenneth A. (1989), Structural Equations with Latent Variables. New York: John Wiley & Sons.

Brown, Steven P. and Robert A. Peterson (1993), "Antecedents and Consequences of Salesperson Job Satisfaction: Meta-Analysis and Assessment of Causal Effects," Journal of Marketing Research, 30 (February), 63-77.

Cascardi, Michele M., Norman G. Poythress, and Alicia A. Hall (2000), "Procedural Justice in the Context of Civil Commitment: An Analogue Study," Behavioral Sciences and the Law, 18 (6), 731-40.

Chonko, Lawrence B., John F. Tanner Jr., and William A. Weeks (1992), "Selling and Sales Management in Action: Reward Preferences of Salespeople," Journal of Personal Selling & Sales Management, 12 (Summer), 67-75.

Churchill, Gilbert A., Jr., Neil M. Ford, and Orville C. Walker Jr. (1979), "Personal Characteristics of Salespeople and the Attractiveness of Alternative Rewards," Journal of Business Research, 7 (1), 25-50.

Colquitt, Jason A., Donald E. Conlon, Michael J. Wesson, Christopher O.L.H. Porter, and K. Yee Ng (2001), "Justice at the Millennium: A Meta-Analytic Review of 25 Years of Organizational Justice Research," Journal of Applied Psychology, 86 (3), 425-45.

Costa, Ana Cristina, Robert A. Roe, and Tharsi Tallieu (2001), "Trust Within Teams: The Relation with Performance Effectiveness," European Journal of Work and Organizational Psychology, 10 (September), 225-44.

Crant, J. Michael and Thomas S. Bateman (1993), "Assignment of Credit and Blame for Performance Outcomes," Academy of Management Journal, 36 (February), 7-27.

Cron, William L., Alan J. Dubinsky, and Ronald E. Michaels (1988), "The Influence of Career Stages on Components of Salesforce Motivation," Journal of Marketing, 52 (January), 78-92.

Dailey, Robert C. and Delaney J. Kirk (1992), "Distributive and Procedural Justice as Antecedents of Job Satisfaction and Intent to Turnover," Human Relations, 45 (March), 305-317.

Denton, Keith D. (1991), "What's Wrong with These Employees?" Business Horizons, 34 (September-October), 45-49.

Dubinsky, Alan J. and Thomas E. Barry (1982), "A Survey of Sales Management Practices," Industrial Marketing Management, 11 (April), 133-41.

Dwyer, Robert F., Paul H. Schurr, and Sejo Oh (1987), "Developing Buyer-Seller Relationships," Journal of Marketing, 51 (April), 11-27.

Farnham, Alan (1989), "The Trust Gap," Fortune, (December 4), 54-74.

Flaherty, Karen E. and James M. Pappas (2000), "The Role of Trust in Salesperson-Sales Manager Relationships," Journal of Personal Selling & Sales Management, 20 (Fall), 271-78.

Folger, Robert A. (1986), "Rethinking Equity Theory: A Referent Cognitions Model," in Justice in Social Relations, H.W. Bierhoff, R.L. Cohen, and J. Greenberg, eds. New York: Plenum Press, 145-62.

----- and Mary A. Konovsky (1989), "Effects of Procedural and Distributive Justice on Reactions to Pay Raise Decisions," Academy of Management Journal, 32 (March), 115-30.

Ford, Neil M., Gilbert A. Churchill Jr., and Orville C. Walker Jr. (1985), "Differences in the Attractiveness of Alternative Rewards Among Industrial Salespeople: Additional Evidence," Journal of Business Research, 13 (April), 123-38.

Futrell, Charles M. and Omer C. Jenkins (1978), "Pay Secrecy Versus Pay Disclosure for Salesmen: A Longitudinal Study," Journal of Marketing Research, 15 (May), 214-19.

Ganesan, Shankar (1994), "Determinants of Long-Term Orientation in Buyer-Seller Relationships," Journal of Marketing, 58 (April), 1-19.

-----, Barton A. Weitz, and George John (1993), "Hiring and Promotion Policies in Salesforce Management: Some Antecedents and Consequences," Journal of Personal Selling & Sales Management, 13 (Spring), 15-26.

George, Jennifer M. (1991), "State or Trait: Effects of Positive Mood on Prosocial Behaviors at Work," Journal of Applied Psychology, 75 (6), 698-709.

Gilliland, Stephen W. (1993), "The Perceived Fairness of Selection Systems: An Organizational Justice Perspective," Academy of Management Journal, 18 (4), 694-734.

Goodson, Jane R. and Gail W. McGee (1991), "Enhancing Individual Perceptions of Objectivity in Performance Appraisal," Journal of Business Research, 22 (June), 293-304.

Goodwin, Cathy and Ivan Ross (1992), "Consumer Responses to Service Failures: Influence of Procedural and Interactional Fairness Perceptions," Journal of Business Research, 25 (September), 149-63.

Greenberg, Jerald (1986), "Determinants of Perceived Fairness of Performance Evaluations," Journal of Applied Psychology, 71 (May), 340-42.

----- (1990), "Organizational Justice: Yesterday, Today, and Tomorrow," Journal of Management, 16 (2), 399-432.

Heneman, H.G. (1974), "Comparisons of Self- and Superior Ratings of Managerial Performance," Journal of Applied Psychology, 59 (October), 638-42.

Iaffaldano, Michelle T. and Paul M. Muchinsky (1985), "Job Satisfaction and Job Performance: A Meta-Analysis," Psychological Bulletin, 97 (March), 251-73.

Ingram, Thomas N. and Danny N. Bellenger (1983), "Personal and Organizational Variables: Their Relative Effect on Reward Valences of Industrial Salespeople," Journal of Marketing Research, 20 (May), 198-205.

Jaworski, Bernard J. and Deborah MacInnis (1989), "Marketing Jobs and Management Controls: Toward a Framework," Journal of Marketing Research, 26 (November), 406-419.

Johnson, Julie T., Hiram C. Barksdale Jr., and James S. Boles (2001), "The Strategic Role of the Salesperson in Reducing Customer Defection in Business Relationships," Journal of Personal Selling & Sales Management, 21 (Spring), 123-34.

Johnston, Mark W., A. Parasuraman, Charles M. Futrell, and William C. Black (1990), "A Longitudinal Assessment of the Impact of Selected Organizational Influences on Salespeople's Organizational Commitment During Early Employment," Journal of Marketing Research, 17 (August), 333-44.

Judge, Timothy A., Carl J. Thoresen, Joyce E. Bono, and Gregory K. Patton (2001), "Job Satisfaction-Job Performance Relationship: A Qualitative and Quantitative Review," Psychological Bulletin, 127 (3), 376-407.

Katz, Daniel and Robert L. Kahn (1978), The Social Psychology of Organizations. New York: John Wiley & Sons.

Konovsky, Mary A. and Russell Cropanzano (1991), "Perceived Fairness of Employee Drug Testing as a Predictor of Employee Attitudes and Job Performance," Journal of Applied Psychology, 76 (October), 698-707.

----- and S. Douglas Pugh (1994), "Citizenship Behavior and Social Exchange," Academy of Management Journal, 37 (3), 656-69.

Korsgaard, M. Audrey, David M. Schweiger, and Harry J. Sapienza (1995), "Building Commitment, Attachment, and Trust in Strategic Decision-Making Teams: The Role of Procedural Justice," Academy of Management Journal, 38 (February), 60-84.

Kumar, Nirmalya, Lisa K. Scheer, and Jan-Benedict E.M. Steenkamp (1995), "The Effects of Perceived Interdependence on Dealer Attitudes," Journal of Marketing Research, 32 (August), 348-56.

Lawler, Edmund E. (1990), Strategic Pay: Aligning Organizational Strategies and Pay Systems. San Francisco: Jossey-Bass.

Leonard, Bill (2001), "Majority of Workers Are Dissatisfied with Pay," HR Magazine, (June), 35-36.

Lester, Scott W., William H. Turnley, James M. Bloodgood, and Mark C. Bolino (2002), "Not Seeing Eye-to-Eye: Differences in Supervisor and Subordinate Perceptions of and Attributions for Psychological Contract Breach," Journal of Organizational Behavior, 23 (February), 39-56.

Leventhal, G.S. (1980), "What Should Be Done with Equity Theory? New Approaches to the Study of Fairness in Social Relationships," in Social Exchange: Advances in Theory and Research, K. Gergen, M. Greenberg, and R. Willis, eds. New York: Plenum Press, 27-55.

Lind, E. Allen and Tom R. Tyler (1988), The Social Psychology of Procedural Justice. New York: Plenum Press.

Livingstone, Linda P., James A. Roberts, and Lawrence B. Chonko (1995), "Perceptions of Internal and External Equity as Predictors of Outside Salespeople's Job Satisfaction," Journal of Personal Selling & Sales Management, 15 (Spring), 33-46.

Lucas, George H., Jr., A. Parasuraman, Robert A. Davis, and Ben M. Enis (1987), "An Empirical Study of Salesforce Turnover," Journal of Marketing, 51 (July), 34-59.

MacKenzie, Scott B., Philip M. Podsakoff, and Michael Ahearne (1998), "Some Possible Antecedents and Consequences of In-Role and Extra-Role Salesperson Performance," Journal of Marketing, 62 (July), 87-98.

Markham, Steven E. (1988), "Pay-for-Performance Dilemma Revisited: Empirical Example of the Importance of Group Effects," Journal of Applied Psychology, 73 (May), 172-80.

Martin, Christopher L. and Nathan Bennett (1996), "The Role of Justice Judgments in Explaining the Relationship Between Job Satisfaction and Organizational Commitment," Group and Organization Management, 21 (March), 84-104.

Martin, David C. and Kathryn M. Bartol (1998), "Performance Appraisal: Maintaining System Effectiveness," Public Personnel Management, 27 (Summer), 223-50.

Masterson, Suzanne S., Kyle Lewis, Barry M. Goldman, and M. Susan Taylor (2000), "Integrating Justice and Social Exchange: The Differing Effects of Fair Procedures and Treatment on Work Relationships," Academy of Management Journal, 41 (4), 738-48.

McDonald, Gerald W. (1981), "Structural Exchange and Marital Interaction," Journal of Marriage and the Family, 43 (November), 825-39.

Miceli, Marcia P. (1993), "Justice and Pay System Satisfaction," in Justice in the Workplace: Approaching Fairness in Human Resource Management, Russell Cropanzano, ed. Hillsdale, NJ: Lawrence Erlbaum Associates, 257-83.

Money, Bruce R. and John L. Graham (1999), "Salesperson Performance, Pay, and Job Satisfaction: Tests of a Model Using Data Collected in the United States and Japan," Journal of International Business Studies, 30 (Spring), 149-69.

Moorman, Christine, Gerald Zaltman, and Rohit Deshpandé (1992), "Factors Affecting Trust in Market Research Relationships," Journal of Marketing, 57 (January), 81-101.

Moorman, Robert H. (1991), "The Relationship Between Organizational Justice and Organizational Citizenship Behaviors: Do Fairness Perceptions Influence Employee Citizenship?" Journal of Applied Psychology, 76 (6), 845-55.

Morgan, Robert M. and Shelby D. Hunt (1994), "The Commitment-Trust Theory of Relationship Marketing," Journal of Marketing, 58 (July), 20-38.

Mowday, Richard T., Richard M. Steers, and Lynn W. Porter (1979), "The Measurement of Organizational Commitment," Journal of Vocational Behavior, 14 (April), 224-47.

Mowen, John C., Janet E. Keith, Stephen W. Brown, and Donald W. Jackson Jr. (1985), "Utilizing Effort and Task Difficulty Information in Evaluating Salespeople," Journal of Marketing Research, 22 (May), 185-91.

Muczyk, Jan P. and Myron Gable (1987), "Managing Sales Performance Through a Comprehensive Performance Appraisal System," Journal of Personal Selling & Sales Management, 7 (May), 41-52.

Netemeyer, Richard G., James S. Boles, Daryl O. McKee, and Robert McMurrian (1997), "An Investigation into the Antecedents of Organizational Citizenship Behaviors in a Personal Selling Context," Journal of Marketing, 61 (July), 85-98.

O'Reilly, C.A. and K.H. Roberts (1974), "Information Filtration in Organizations: Three Experiments," Organizational Behavior and Human Performance, 11 (April), 253-65.

Organ, Dennis W. (1977), "A Reappraisal and Reinterpretation of the Satisfaction Causes Performance Hypothesis," Academy of Management Review, 2 (January), 46-53.

----- (1988), "A Restatement of the Satisfaction-Performance Hypothesis," Journal of Management, 14 (4), 547-57.

Pettijohn, Charles E., Linda S. Pettijohn, and Albert J. Taylor (2000), "An Exploratory Analysis of Salesperson Perceptions of the Criteria Used in Performance Appraisals, Job Satisfaction, and Organizational Commitment," Journal of Personal Selling & Sales Management, 20 (Spring), 77-80.

Petty, M.M., Gail W. McGee, and Jerry W. Cavender (1984), "A Meta-Analysis of the Relationship Between Individual Job Satisfaction and Individual Performance," Academy of Management Review, 9 (October), 712-21.

Podsakoff, Philip M. and Scott B. MacKenzie (1994), "Organizational Citizenship Behaviors and Sales Unit Effectiveness," Journal of Marketing Research, 31 (August), 351-63.

Ramaswami, Sridhar N. (1996), "Marketing Controls and Dysfunctional Employee Behaviors: A Test of Traditional and Contingency Theory Postulates," Journal of Marketing, 60 (April), 105-120.

-----, Srini S. Srinivasan, and Stephen A. Gorton (1997), "Information Asymmetry Between Salesperson and Supervisor: Postulates from Agency and Social Exchange Theories," Journal of Personal Selling & Sales Management, 17 (Summer), 29-50.

Rich, Gregory A. (1997), "The Sales Manager as a Role Model: Effects on Trust, Job Satisfaction, and Performance of Salespeople," Journal of the Academy of Marketing Science, 25 (October), 319-28.

Roberson, Quinetta M., Neta A. Moye, and Edwin A. Locke (1999), "Identifying a Missing Link Between Participation and Satisfaction: The Mediating Role of Procedural Justice Perceptions," Journal of Applied Psychology, 84 (4), 585-93.

Sarin, Shikar and Vijay Mahajan (2001), "The Effect of Reward Structures on the Performance of Cross-Functional Product Development Teams," Journal of Marketing, 65 (April), 35-53.

Sashkin, Marshall and Richard L. Williams (1990), "Does Fairness Make a Difference?" Organizational Dynamics, 19 (2), 56-71.

Singh, Jagdip, Willem Verbeke, and Gary K. Rhoads (1996), "Do Organizational Practices Matter in Role Stress Processes? A Study of Direct and Moderating Effects for Marketing-Oriented Boundary Spanners," Journal of Marketing, 60 (July), 69-86.

Summers, Timothy P. and William H. Hendrix (1991), "Modeling the Role of Pay Equity Perception: A Field Study," Journal of Occupational Psychology, 64 (June), 145-57.

Teas, Kenneth R. (1981), "Selling Task Characteristics and the Job Satisfaction of Industrial Salespersons," Journal of Personal Selling & Sales Management, 1 (Spring-Summer), 18-26.

Thibaut, John W. and Laurens Walker (1975), Procedural Justice: A Psychological Analysis. Hillsdale, NJ: Lawrence Erlbaum Associates.

Turban, Daniel B. and Allan P. Jones (1988), "Supervisor-Subordinate Similarity: Types, Effects, and Mechanisms," Journal of Applied Psychology, 73 (2), 228-34.

Tyagi, Pradeep K. (1985), "Relative Importance of Key Job Dimensions and Leadership Behaviors in Motivating Salesperson Work Performance," Journal of Marketing, 49 (Summer), 76-86.

Tyler, Tom R. (1989), "The Psychology of Procedural Justice: A Test of the Group-Value Model," Journal of Applied Psychology, 57 (5), 830-38.

Vroom, Victor H. (1964), Work and Motivation. New York: John Wiley & Sons.

Williams, Larry J. and John T. Hazer (1986), "Antecedents and Consequences of Satisfaction and Commitment in Trust Models: A Reanalysis Using Latent Variable Structural Equation Methods," Journal of Applied Psychology, 71 (2), 219-23.

Unless otherwise noted, we measured the following items on a five-point Likert scale where 1 = "strongly disagree" and 5 = "strongly agree." The items marked with [O] were removed from the analyses because of poor internal consistency with their respective scales.

Linkage to Rewards (Folger and Konovsky 1989)

1. My pay increases are based upon how my performance compares with my goals.

2. My merit increases are directly tied to my performance.

Consistent/Unbiased Application of Performance Standards (Tyler 1989)

1. Do you think that you received a better or worse merit increase than others because of your race, sex, age, nationality, or some other characteristic of you as a person? (Three-point scale: 1 = "yes"; -1 = "no"; 0 = "don't know")

2. Do you think that your supervisor treated you worse than others because of your race, sex, age, nationality, or some other characteristic of you as a person? (Three-point scale: 1 = "yes"; -1 = "no"; 0 = "don't know")

Performance Improvement Plan (Folger and Konovsky 1989)

Indicate the extent to which you believe your supervisor did each of the following during the last performance cycle:

1. Discussed plans or objectives to improve your performance.
2. Asked for your ideas on what you could do to improve your performance.
3. Developed an action plan for future performance.

Measure Appropriateness (New Scale)

1. My performance is evaluated on all relevant and important skill areas of my job.

• 2. Standards or performance targets are set for each skill area of my job.
• 3. The targets set for my job, I believe, are appropriate.
• 4. The standards used in performance review are all based on a thorough analysis of the job I perform.

Participation (Teas 1981; Vroom 1964)

1. I am allowed a high degree of influence in the determination of my work goals.

• 2. I really have little voice in the formulation of my work goals. [reverse scored]
• 3. The setting of my work goals is pretty much within my control.
• 4. My supervisor usually asks for my opinions and thoughts when determining my work goals.

Distributive Fairness (Folger and Konovsky 1989)

1. I consider the size of my last merit increase to be fair.

• 2. My last merit increase gave me the full amount I deserved.
• 3. The size of my last merit increase was more than what I expected.
• 4. The level of merit increase I received was (1 = "very unfair"; 4 = "very fair").

Procedural Fairness (Folger and Konovsky 1989; Moorman 1991)

1. How much of a chance or opportunity did your supervisor give you to describe your achievements and contributions to him/her before making your merit increase decision? (Four-point Likert scale: 1 = "a great deal of opportunity"; 4 = "not much opportunity at all")

• 2. How much influence did you have over the merit decision made by your supervisor? (Four-point scale: 1 = "a great deal of influence"; 4 = "not much influence at all)
• 3. How much consideration did your supervisor give to what you said when making merit increase decisions? (Four-point scale: 1 = "a great deal of consideration"; "4 = "not much consideration at all)
• 4. Overall, how fair were the methods used by your supervisor to make your merit increase decision? (Four-point scale: 1 = "very fair"; 4 = "very unfair")

Interactional Fairness (Folger and Konovsky 1989; Moorman 1991)

Indicate the extent to which you believe your supervisor did each of the following during the last performance cycle:

1. Was honest and ethical in dealing with you.
2. Showed a real interest in trying to be fair.
3. Treated you with respect and dignity.
4. Was sensitive to your personal needs.
5. Showed concerns for your rights as an employee.

Supervisory Trust (O'Reilly and Roberts 1974)

1. How free do you feel to discuss with your immediate supervisor the problems and difficulties in your job without jeopardizing your position or having it held against you later? (Seven-point scale: 1 = "completely free"; 7 = "very cautiously")

• 2. Immediate supervisors at times must make decisions which seem to be against the interest of employees. When this happens to you as a employee, how much trust do you have that your immediate supervisor's decision was justified by other considerations? (Seven-point scale: 1 = "trust completely"; 7 = "feel very distrustful")
• 3. To what extent do you have trust and confidence in your immediate supervisor regarding his or her general fairness? (Seven-point scale: 1 = "completely"; 7 = "very little")

Job Satisfaction (Lucas et al. 1987)

1. My job is satisfying.

• 2. My job is exciting.
• 3. I'm really doing something worthwhile in my job.
• 4. The work I perform gives me a sense of accomplishment.

Job Performance (New Scale)

This employee

1. Performed above average on annual sales objective.

• 2. Performed above average on business growth objective.
• 3. Performed above average on professional growth objective.
• 4. Is one of our best managers.
• 5. Is outstanding.

Opportunistic Behaviors (Jaworski and MacInnis 1989)

1. I sometimes tend to ignore certain job-related activities simply because they are not monitored by our supervisors.

• 2. I sometimes work on unimportant activities simply because they are evaluated by our supervisors. [O]
• 3. When my performance is inconsistent, I have tried to make it appear consistent.
• 4. When presenting data to upper management, I generally try to emphasize data that reflects favorably upon me.
• 5. Most employees underreport sales potential in their territory to obtain lower sales targets. [O]
• 6. When presenting data to upper management, I generally try to avoid being the bearer of bad news.

Organizational Commitment (Mowday, Steers, and Porter 1979)

1. I am willing to put in a great deal of effort beyond that normally expected to help this organization be successful.

• 2. I talk about this organization to my friends as a great place to work.
• 3. I would accept almost any type of job assignment in order to keep working for this organization.
• 4. I find that my values and the organization's values are similar. [O]
• 5. I am proud to tell others that I am part of this organization.
• 6. I am extremely glad that I chose this organization to work over others I was considering at the time I joined.
• 7. This organization really inspires the very best in me in the way of job performance. [O]
• 8. I really care about the fate of this organization.
• 9. For me, this is the best of all organizations for which to work. [O]

~~~~~~~~

By Sridhar N. Ramaswami and Jagdip Singh

Sridhar N. Ramaswami is Professor of Marketing, Iowa State University.

Jagdip Singh is Professor of Marketing, Weatherhead School of Management, Case Western Reserve University.

Antecedents and Consequences of the Service Climate in Boundary-Spanning Self-Managing Service Teams 

In this article, the authors examine antecedents and consequences of the service climate in boundary-spanning self-managing teams (SMTs) that deliver financial services. Using data from members of 61 SMTs and their customers, the authors show a differential impact of the SMT service climate on various marketing performance measures. Furthermore, they obtain support for independent group-level effects of intrateam support and team member flexibility on employee perceptions of the SMT service climate. Both effects are persistent over time and demonstrate that collective perceptions in the SMT have incremental value in the explanation of the service climate.

Research on boundary-spanning service employees has shown that delegation of authority to the front line allows for greater flexibility and adaptability in the performance of service activities through better problem solving, closer employee cooperation, and more efficient knowledge transfer (Hartline and Ferrell 1996). In apparent recognition of this, some companies have organized their front-office operations around self-managing teams (SMTs), or groups of interdependent employees that have the collective authority and responsibility to manage and perform relatively whole tasks. Team members are typically cross-trained in various skills, including developing work routines, planning, and monitoring team performance (Yeatts and Hyten 1998). Companies such as Charles Schwab, Taco Bell, Prudential, Pacific Bell (now part of SBC Communications), CIGNA, Welch Foods, and Xerox have implemented boundary-spanning SMTs (Batt 1999; Cameron and Boise 1995; Wageman 1997).

It has been argued that self-management is an excellent mechanism for improving the performance of the employee-customer interface (Gilson, Shalley, and Blum 2001). However, this claim is not substantiated by the empirical evidence. Whereas Batt (1999) shows that front-office service SMTs perform significantly better in terms of service quality and sales volume than do teams under management control, Chaston (1998) reports an adverse impact on service quality and productivity. Whereas Wageman (1997, p. 32) states that the performance of field-service SMTs at Xerox is "critical to the company's ultimate success," other reports from the business press contest the validity of this assumption (e.g., Zemke 1993). Empirical inconsistencies, conflicting anecdotes, and a lack of theoretical development regarding SMTs in service settings emphasize the need for research that addresses four important theoretical and empirical issues that have been unresolved in previous studies.

First, the research to date has virtually ignored the development of a mediating construct to account convincingly for the apparent inconsistencies in SMT performance. At the firm level, Schneider, White, and Paul (1998) demonstrate that a service climate is a key mediating factor in the prediction of marketing performance. Accordingly, we advance a construct of the SMT service climate that is proximal to perceptions of work-group practices to explain performance variability among teams.

Second, previous research on customer-contact SMTs has focused predominantly on processes within the team (e.g., Cohen, Chang, and Ledford 1997). However, as Hackman (1992) argues, teams do not operate in isolation. Therefore, climate perceptions may also be the result of organizational context characteristics. We develop a conceptual framework that takes into account both team-and company-related predictors of the SMT service climate.

A third issue that has not been explored is whether climate-defining team characteristics have an impact at the work-group level that is beyond the perception of individual employees. Each SMT may develop a unique set of shared perceptions of desirable behavior (e.g., the level of support to other team members), and between-group differences may be contingent on these perceptions (Mathieu and Kohler 1990). Therefore, we examine whether group-level aggregations of team member perceptions incrementally determine employee perceptions of the SMT service climate. The group-level effects may be contingent on the type of service delivery. Stewart and Barrick (2000) demonstrate that task type moderates variability in the magnitude of reported predictor-criterion relationships in explaining manufacturing SMT performance. We extend this finding by considering distinct types of service delivery in our analysis of group-level antecedents and consequences of the SMT service climate.

Fourth, the impact of boundary-spanning teams on business performance measures (Batt 2002) is unknown. Part of the gap in the knowledge about SMT effectiveness stems from the complexity in considering various types of performance measures (e.g., customer evaluations versus productivity measures) across different types of service delivery (Batt 2002). Another reason is that extant studies may not have been able to provide definitive conclusions on frontline SMT effectiveness because they employed cross-sectional data on relationships that might have needed to be separated in time (Griffin 1991). Therefore, we examine the lagged effects of the SMT service climate on customer perceived quality, share of customer, and sales productivity for two types of service delivery.

The increasing importance and ubiquity of boundary-spanning SMTs, despite inconsistent and sometimes contradictory findings, create the need for more definitive research that describes and defines the nature and scope of SMTs and their possible effects. We propose four additions to SMT research and test the theoretical and empirical advantages and implications of these additions.

SMT Service Climate

Although various employee-based measures (e.g., job satisfaction) have been advanced as drivers of service performance, it also has been argued that service climate has superior predictive power (Schneider, Wheeler, and Cox 1992). We extend previous research by adopting a team-level focus on service climate. Our conceptual point of departure for developing several definitional assumptions is Katz and Kahn's (1978) description of climate being the result of a distinct pattern of individual team members' collective beliefs developed through members' interaction with their social environments. First, this description theoretically relates climate to antecedent variables in the organizational and team contexts (Lindell and Brandt 2000). Second, Katz and Kahn (1978) posit that perceptions of climate are essentially a property of the individual member that can be aggregated to reflect a group-level construct (see James and James 1989). Third, as climate is related to various environments, different climates may exist for organizational goals and structural levels. Proximal measures that conceptualize climate in terms of both goals (e.g., customer service) and levels (e.g., the team) produce strong relationships with targeted performance parameters (Tesluk et al. 1995). Finally, because climate involves the construction of shared meaning through the process of interaction, the process is dynamic and in line with Hackman's (1987) process criterion of effectiveness, which relates to team members' effort, knowledge, skill, and performance in achieving team goals. Thus, we define SMT service climate as the collective beliefs of SMT members on effort, knowledge, skills, and performance with regard to effective service delivery.

Antecedents of SMT Service Climate: An Individual-Level Perspective

Our framework, which relates SMT service climate to its antecedents, has two distinct conceptual roots: ( 1) Hackman's (1987) normative model of team effectiveness, which distinguishes intra-and extrateam factors, and ( 2) the involvement approach (Bowen and Lawler 1992), according to which employees are given the authority and resources to coordinate, plan, and control the service delivery process. Three main characteristics that differentiate service SMTs from other traditional work groups govern our choice of predictor variables: ( 1) higher levels of autonomy, ( 2) functional flexibility, and ( 3) interdependency within and between teams (Campion, Medsker, and Higgs 1993). In general, SMTs are designed with a certain degree of role-related diversity (Yeatts and Hyten 1998). Researchers have posited that perceptions of collective phenomena (i.e., service climate) represent cognitive interpretations of proximal structures and processes based on a person's experience, values, knowledge, and expertise (Brown and Leigh 1996). Thus, members of the same SMT may have different perceptions of the SMT service climate defining antecedents. Prior research on work groups has demonstrated that within-group perceptual deviation reflects systematic (not random) variance that may represent differential cognitive appraisals of the team environment (Van Yperen and Snijders 2000). Thus, we postulate predictor-criterion relationships at the individual level.

Hackman (1987) posits that a supportive organizational context is a major determinant of group effectiveness. A central component of the involvement approach is empowerment, which refers to the notion that service employees must be given a certain degree of autonomy and be able to perform job-related activities with skill (Hartline and Ferrell 1996). Recent research on production teams demonstrates that senior management's lack of tolerance for self-management may be an important barrier to SMT effectiveness (Balkema and Molleman 1999). Furthermore, in the case of boundary-spanning SMTs, it has been consistently argued that perceived autonomy is critical to the attitude and behavior of customer-contact personnel (Batt 1999; Van Mierlo et al. 2001; Wageman 1995). Thus:

H₁: Tolerance for self-management positively affects employees' perceptions of the SMT service climate.

In addition, SMT members need to be able to use delegated authority optimally; this ability has been associated with several synergistic processes in teams (Hackman 1987). The SMT members should be capable of performing various team tasks, whether operational, managerial, or administrative (Spreitzer, Cohen, and Ledford 1999). As the spectrum of SMTs' tasks grows, job assignments rapidly evolve, SMTs need to become highly interdependent, and there is a need for flexible and multiskilled members (Batt 1999; Sundstrom, de Meuse, and Futrell 1990). Recent research (Marks, Mathieu, and Zaccaro 2001; Mathieu et al. 2000) has demonstrated that the ability to perform interrole behaviors facilitates attainment of team goals. We hypothesize that perceptions of team member flexibility form another relevant foundation for the establishment of service climate:

H₂: Flexibility of team members positively affects employees' perceptions of the SMT service climate.

The literature on employee involvement has demonstrated that coworker involvement reduces perceptions of boundary-spanner role stress and increases service performance (Bettencourt and Brown 1997). The implication is that when service employees experience peer-based learning and coworkers' service-driven attitude, they will be motivated to carry over this attitude to their customer encounters. Frequently, cooperative interaction within and between SMTs is required to address customer service requests successfully and to create a service-oriented work environment (Horwitz and Neville 1996). The services marketing literature posits that mutual support among employees is essential to the implementation of service-quality improvements (Berry, Parasuraman, and Zeithaml 1994). Batt (1999) has shown that collaborative endeavors are a key success factor of SMT effectiveness in boundary positions. Conceptually, we distinguish between inter-and intrateam support. Interteam support refers to team member perceptions of the internal service and communication between teams and other units in the organization, whereas intrateam support pertains to team members' willingness to offer help and to deliver service to other members of the group in order to attain work-group goals (Campion, Medsker, and Higgs 1993). We hypothesize the following:

H₃: Employee perceptions of the SMT service climate are positively affected by (a) interteam support and (b) intrateam support.

Although some researchers claim that the individual perspective constitutes the only proper unit of analysis, others argue that the study of group phenomena can be analyzed in a meaningful way only at the group level (Lindell and Brandt 2000). Therefore, in the next section, we discuss the conceptualization of the aforementioned predictor variables at the group level.

Antecedents of SMT Service Climate: A Group-Level Perspective

It has been advanced that by evoking the notion of climate, each team can develop a unique set of norms and mental models regarding desirable behavior (e.g., the level of support to other team members), thus reflecting between-group differences (Mathieu and Kohler 1990). Studies on shared mental models (Mathieu et al. 2000) and transactive memory in teams (Liang, Moreland, and Argote 1995) demonstrate that team members develop shared beliefs about their team that instigate team members to develop interrelated knowledge and norm structures to facilitate group processes. Lindell and Brandt (2000, p. 332) state that shared beliefs reflect that individual members "are socialized to act in similar ways, are exposed to similar features within contexts, and come to share their interpretations with others in the setting." Beliefs are conceptually distinct from constructs that exist at the group level only (e.g., team size) (Gully et al. 2002). Aggregate-level constructs reflect psychosocial traits that are not captured by the individual-level measurement (Hackman 1992). These constructs may differentially influence individual members' perceptions of the SMT service climate. Group-level constructs strongly reflect the basic assumption of synergistic processes within the SMT (Hackman 1987). Poole and McPhee (1983, p. 213) view sources of work-unit climate as collective attitudes that are "continually produced and reproduced by members' interactions." Particularly in service SMTs, there is a relatively high level of interdependence and interactions (Batt 1999), in which shared perceptions, unique to the work unit, are formed.

To better understand the similarities and differences inherent in multiple-level constructs, scholars have used the typology of elemental composition (Bliese 2000; Chan 1998). Elemental composition takes place when a higher-level construct consists of collective lower-level measures. The composition model for the SMT antecedents in our study is the direct consensus model, which characterizes climate-defining antecedents as properties of the individual team member. At the same time, when there is consensus between individual perceptions (e.g., on intrateam support), the aggregate composes a construct at the work-group level that represents shared perceptions of a collective belief. Prior research on teams' withdrawal behavior (absenteeism and lateness) in boundary-spanning service settings provides evidence of the influence of group-level predictors on individual employee behavior beyond individual-level antecedents. Mathieu and Kohler (1990) and Blau (1995) report significant effects of aggregated group-level variables on individual employee behavior in auto repair, hospital, and banking service teams. Their results seem to support Bryk and Raudenbush's (1992) contention that group-level aggregations of contextual properties represent a distinct perspective that may not be captured by individual-level measures. Therefore, we hypothesize the following:

H₄: At the group level of analysis, the positive effects of (a) tolerance for self-management, (b) team member flexibility, (c) interteam support, and (d) intrateam support account for a significant amount of additional variance in individual employees' perceptions of the SMT service climate.

Antecedent-SMT Service Climate Relationships Across Service Types

Positive linear relationships between SMT characteristics and member-related outcomes (e.g., employee satisfaction, organizational commitment) within and across various frontline service settings have not yielded a consistent pattern (Batt 1999; Cohen, Chang, and Ledford 1997; Gilson, Shalley, and Blum 2001; Wageman 1997). Variability in the magnitude of reported predictor-criterion relationships may indicate the presence of moderator variables, such as task characteristics. In a manufacturing setting, Stewart and Barrick (2000) show that SMTs that are responsible for routine tasks are less likely to be affected by flexibility and interdependency than are SMTs that perform nonroutine activities. Recent research by Marks and colleagues (2002) demonstrates that the development of shared beliefs regarding team processes fosters coordinated task behavior and the ability to adapt dynamically in uncertain environments. Particularly in the case of nonroutine tasks, norm congruence and shared understanding on how to function as an SMT affect climate perceptions. Because services have also been classified in terms of a routine-nonroutine continuum (Davis 1999), we expect service type to moderate the impact of group-level predictor variables. We posit the following:

H₅: Service type moderates the positive effects of (a) tolerance for self-management, (b) team member flexibility, (c) interteam support, and (d) intrateam support on employees' perceptions of the SMT service climate, such that effects are significantly stronger at the group level for nonroutine services than for routine ones.

Consequences of SMT Service Climate

There is ample evidence that employee perceptions of service climate at the firm level have a positive influence on customer perceptions of service quality (Schneider et al. 1996; Schneider and Bowen 1985; Schneider, White, and Paul 1998). In addition to psychological outcome parameters, it has been argued that the behavioral outcome "share of customer," or the number of services purchased from a specific service provider, is a key marketing-performance indicator (Babin and Attaway 2000). The overall premise is that the impact of policies and practices aimed to serve the customer should also be observable in customer behavior, because both perceived service quality and purchase behavior are closely related and should be evaluated simultaneously when pursuing profitability (Soteriou and Zenios 1999). Service firms may also focus on productivity as a performance parameter, specifically on quantifiable behavioral standards of employee behavior, such as volume of services sold (Singh 2000). However, recent evidence suggests that, especially in the area of services, firms should make trade-offs between establishing a focus on delivering quality services and sales ratios per employee (Anderson, Fornell, and Rust 1997; Singh 2000). Thus, we expect that a strong SMT service climate with an emphasis on providing high-standard services will result in decreased productivity. It has been suggested in both the service marketing and the team literature that inconsistencies in performance assessment are contingent on the time frame used in the analysis (Griffin 1991). Empirical evidence from the services marketing literature shows that customers' assessments of service quality are relatively constant and subject to slow change (Bolton and Drew 1991). Moreover, Bernhardt, Donthu, and Kennett (2000) argue that the true impact of service changes in terms of observable customer behavior (e.g., actual services purchased) can be assessed only longitudinally. Finally, Griffin (1991) reports a delay between the implementation of frontline service SMTs and performance improvements. Therefore, we propose the following:

H₆: At the group level of analysis, SMT service climate at T1 positively affects (a) customer perceived service quality at T2 and (b) share of customer at T2.

H₇: At the group level of analysis, SMT service climate at T1 negatively affects sales productivity at T2.

Assessment of Consequences Across Service Types

Finally, we posit that the impact of SMT service climate on the aforementioned parameters may depend on the type of service. Nonroutine service delivery frequently involves dealing with complex problems and equivocal situations that may benefit from an extended dialogue within the team and between team members and customers. We expect that a climate that fosters service excellence is more influential in such circumstances:

H₈: At the group level of analysis, service type moderates the positive effects of SMT service climate at T1 (a) on customer perceived service quality at T2 and (b) on share of customer at T2 and moderates (c) the negative effect on sales productivity at T2, such that effects are significantly stronger at the group level for nonroutine services than for routine ones.

Figure 1 reflects our conceptual framework and provides an overview of the issues discussed thus far.

Research Setting

Members of a large Dutch bank's SMTs and customers were surveyed. The bank employs approximately 48,000 people and has 424 branch offices. It operates in both business and consumer markets and promotes service excellence as a key to marketing success. In each branch, separate routine and nonroutine SMTs are responsible for servicing consumers, and other SMTs deal with business customers or are responsible for internal operations (e.g., human resources management, general and technical services). In our study, we focused on SMTs that deliver consumer services. Each branch offers a wide range of services to consumers: nonroutine, knowledge-intensive services (e.g., investment consulting, trust services, estate planning) and routine, transaction-intensive services (e.g., checks and deposits, currency exchanges, credit application accounts). At each branch, separate front-office SMTs are responsible for the different service types (e.g., financial versus client advisory teams). The bank considered implementation of team-based self-management an important organizational change process. Therefore, the practical rationale for conducting our study at the initiation phase (T1) and seven months later (T2) was to evaluate the central role of SMT service climate and examine its impact on service performance.

Sampling and Surveying

Of 848 boundary-spanning SMTs, we randomly selected a sample of 100. Unwillingness to cooperate and the lack of sales productivity and customer data resulted in a total set of 61 SMTs. We collected data from individual employees organized in SMTs using self-report questionnaires (at T1 and at T2) and from their customers (at T2) using mail questionnaires. For the employee survey, we invited all members of the SMT to participate. In total, 939 questionnaires were returned at T1 (76.4%) and 730 questionnaires at T2 (62.1%). For the customer survey, we drew a random sample of 150 customers per SMT at T2. In total, 1884 questionnaires were returned (20.8%). For the employee survey, we used 568 questionnaires from 36 nonroutine service teams (316 from 36 teams at T1 and 252 from 33 teams at T2) and 917 questionnaires from 25 routine service teams (509 from 25 teams at T1 and 408 from 23 teams at T2) for further analysis. For the customer survey, we analyzed 957 questionnaires of customers of 36 nonroutine service teams and 577 questionnaires of 25 routine service teams. With respect to the customer survey, we acquired the following sample profile: Most respondents were male (63.8%) and older than 44 years of age (58.5%). Virtually all respondents had a long-lasting relationship with the bank (93.1% more than five years) and half of them visited the bank at least once a month (49.8%). Detailed background data on participants at T1 as well as functional areas and responsibilities of the SMTs are included in Table 1.

Measurement Issues

The assessment of the SMT service climate involved items we specifically developed for this study; in-depth interviews with frontline employees; and scales developed by Schneider, Wheeler, and Cox (1992) and Peccei and Rosenthal (1997). We based the operationalization of tolerance for self-management (seven items) largely on the tolerance-of-freedom instrument of the Leader Behavior Description Questionnaire, which assesses the degree of autonomy given to employees to manage their task responsibilities themselves (Cook et al. 1981). We measured team member flexibility and inter-and intrateam support with scales developed by Campion, Medsker, and Higgs (1993). We measured all scale items for the employee survey on a seven-point scale that ranged from 1 = "strongly disagree" to 7 = "strongly agree." A qualified translator translated items into Dutch using the double back-translation procedure (Brislin 1980).

We employed two techniques to test the factor structure and the item loadings of the data collected at T1. We initially examined coefficient alphas and the factor structure (using principal components analysis) for all the scale items simultaneously. We achieved a five-factor structure in which items loaded on a priori dimensions. In addition, we conducted a confirmatory factor analysis (CFA) using LISREL to assess the measurement properties of the items. The fit indexes of the proposed factor model, construct reliabilities of the scales, and confirmatory factor loadings with t-values for each item are reported in Table 2. The indexes of the proposed factor model provide a good fit and show unidimensionality of the scales.( n1) We tested construct reliabilities of the scales by means of Cronbach's alpha. Coefficients of all measures were greater than .65, which implies that reliability is acceptable (Nunnally and Bernstein 1994).

Next, we examined within-method convergent validity by investigating the significance and magnitude of the item loadings. All items loaded significantly on their respective construct (minimum t-value = 12.50), and 95% of all items had a standardized loading of at least .50. In addition, we evaluated discriminant validity. All chi-square difference tests (1 degree of freedom [d.f.]) were significant (p < .05), which indicates that all pairs of constructs correlated at less than one. Subsequently, we included as control variables the demographic variables education, organizational tenure, and age, as well as the work-group-design variables team size and percentage of front-office activities (i.e., direct customer-contact responsibilities) and nonroutine services, which is a dummy for service type (we coded nonroutine services as one and routine services as zero).

We measured two performance measures, customer perceived service quality and share of customer, using the customer survey. We based the scale for service quality on the SERVQUAL instrument developed by Parasuraman, Zeithaml, and Berry (1988). Our goal was to measure the quality of the services delivered by the service employees. Therefore, we restricted our measure to eight items that specifically addressed employee-related aspects of quality (Hartline and Ferrell 1996). We measured items on a five-point scale that ranged from 1 = "very dissatisfied" to 5 = "very satisfied." Principal components analysis showed construct validity (we extracted a single factor; loadings for all items were greater than .70). In addition, we conducted CFA to assess measurement properties. The fit indexes are reported in Table 2; they demonstrate unidimensionality, construct reliability, and convergent validity of the construct. To obtain a measure that we call "share of customer," customers were asked to indicate which types of services they used. The share-of-customer score for an SMT pertains to a group mean based on the average number of different service categories used by the customers of a specific team (see Table 2). Finally, we used data from the bank's internal database on the two service types (i.e., life insurance policies and investment portfolios for nonroutine services and number of checking and savings accounts for routine services). These parameters reflect the average amount of services sold per team member per year; we used them as a measure of sales productivity.

Means, standard deviations, and individual-and grouplevel correlations between the employee variables are presented in Table 3, which reveals that nonroutine service delivery is associated with higher-educated employees, smaller team sizes, higher percentages of front-office workers, and higher levels of tolerance for self-management than is routine service delivery. In Table 4, group-level means, standard deviations, and (partial) correlations of employee variables and external outcomes are presented. The SMT service climate appears to have the highest correlations with customers' perceived service quality. Furthermore, the correlations between antecedents and perceived service quality are noticeably weaker when the effect of SMT service climate is accounted for, which implies that the SMT climate mediates the relationships between antecedents and perceived service quality (see Baron and Kenny 1986). Compared with share of customer and sales productivity, the mediating role of SMT service climate is less obvious.

Analysis

To justify data aggregation, we examined tolerance for self-management, flexibility of team members, inter-and intrateam support, and SMT service climate on within-group agreement and interdependence. First, the average rWG(j) coefficients (ranging from .84 to .95 at T1 and at T2) are high for all variables and show high ratings consistency among employees within groups (James 1982). Second, the intraclass correlation (ICC) ( 1) coefficients (ranging from .06 to .18 at T1 and from .05 to .14 at T2) indicate that for all variables, a small to moderate part concerns between-group variance. To determine interdependence appropriately, it is also relevant to consider group size (Bliese 2000). Therefore, we calculated ICC ( 2) coefficients, because this measure of interdependence accounts for group size. Of the ICC ( 2) values, 90% are greater than .50, which provides evidence for reliable group means and enables the detection of group-level relationships, even in the case of relatively small ICC ( 1) values (Bliese 2000).

We tested H₁-H₅ with employee data at T1 and estimated multilevel models using MLwiN software (Rasbash et al. 2000). We first included the control variables and the antecedents (Model A1). We then specified interactions between the dummy variable nonroutine services and the antecedents (Model A2).( n2) Model specification is provided in the Appendix.

To compare individual-and group-level effects of the antecedents on SMT service climate, we split the antecedent variables into the group mean and within-group deviation score (i.e., individual score - group mean).( n3) The coefficient of the group means reflects the group-level effect, whereas the coefficient of the within-group deviation scores reflects the individual-level effect (Bryk and Raudenbush 1992; Snijders and Bosker 1999). To interpret the coefficients appropriately, we needed to know whether the individual-and group-level coefficients were equal. First, when both coefficients are significant and equal, the variable functions principally at the individual level, and there is no separate main effect at the group level. Second, when only the individual-level coefficient is significant and the group-level coefficient is not significant, the effect is solely based on social comparison within the group. Finally, when both the group-and the individual-level coefficients are significant but differ in magnitude, there is both an individual-level effect and an independent group-level effect (Van Yperen and Snijders 2000).

Table 5 contains the results of the multilevel analyses. There are positive individual-level effects of tolerance for self-management, flexibility, and inter-and intrateam support on the SMT service climate, in support of H₁-H₃. In addition, we tested the cross-level hypothesis (H4). A comparison of the individual-and group-level coefficients shows that tolerance for self-management and flexibility have significant, positive group-level effects on the SMT service climate. However, these do not significantly differ from the individual-level effects. Thus, the results indicate no support for H_4a and H_4b. The group-level effect of interteam support is nonsignificant and significantly weaker than its individual-level counterpart. Therefore, H_4c is rejected. Only intrateam support shows a significant, positive group-level effect on SMT service climate that is significantly stronger than its individual-level effect, in support of H_4d.

With respect to the control variables, our findings show a significant, positive group-level effect of age on SMT service-climate perceptions that is significantly stronger than its individual-level counterpart. Similarly, we find a significant, negative group-level effect of organizational tenure on the outcome variable. At the same time, however, although the group-level predictors are highly positively correlated, each has a negligible correlation with the SMT service climate. Yet the inclusion of group-level tenure as a second predictor of the SMT service climate in addition to group-level age significantly increases the model fit (Χ²(l) = 5.68; p < .05) and results in significant effects of both predictors on the SMT service climate. In contrast, inclusion of age as a second predictor beyond tenure results in a better model fit (Χ²(l) = 5.56; p < .05) and significant effects for the two predictors. Therefore, it appears that both tenure and age can be considered suppressor variables that mutually suppress each other's irrelevant variance for the prediction of the SMT service climate. Essentially, suppressors partial out the variance in the predictor variable that is due to measurement artifacts (Hinkle, Wiersma, and Jurs 1994). Therefore, to determine the effects of tenure and age on SMT service climate adequately, it is necessary to include both variables in the model. Finally, the results show that nonroutine service tasks have a positive impact on climate perceptions.

In addition, we tested whether the relationships between antecedents and the SMT service climate differ across service settings by specifying interactions between nonroutine services and antecedents. The inclusion of the interaction terms does not significantly improve the fit of Model A2 compared with Model A1. We found one significant, positive interaction of nonroutine services x team member flexibility at the group level. The addition of this interaction term to the model leads to a significant increase in model fit (Χ²(l) = 5.92; p < .05), indicating support for H_5b. Specifically, the magnitude of this group-level interaction effect is significantly stronger than its individual-level counterpart.

The specified group-level interactions of nonroutine services with tolerance for self-management and inter-and intrateam support appear to be nonsignificant. This implies that there are no differences in group-level effects of the antecedents across service types. Thus, H_5a, H_5c, and H_5d are rejected. Finally, the percentage of explained variance at the grouplevel is higher than that at the individual level, indicating that between-group differences of SMT service climate are more effectively explained than within-group differences.

To substantiate our results, we performed additional analyses. The results of the analyses are presented in Table 6. First, we conducted a lagged analysis and investigated the delayed effects of the antecedent variables at T1 on SMT service climate at T2 (Model B1). To control for an employee's previous evaluation of SMT service climate at T1, we included this variable as an additional predictor (see Bolton and Drew 1991). The reported estimates reveal a somewhat weaker but largely consistent pattern as compared with the cross-sectional analysis (Model A2). Specifically, these findings empirically demonstrate that the antecedents have a positive impact on the development of the SMT service climate over time. Second, to test the accuracy and stability of parameter estimates (Model B2), we conducted the simulation procedure using Gibbs sampling (Gilks, Richardson, and Spiegelhalter 1996).( n4) The parameter values we obtained through this simulation were fairly similar to the Model A2 estimates. Third, we performed the Chow F_c test to determine whether there had been a structural change in the antecedent-SMT service climate relationships between T1 and T2. The Chow F_c test reported no significant F value (F_{25, 1435} = 1.00; p = .46), which indicates that there were no structural alterations over time. Overall, this model validation indicates insensitivity of the results to various time frames and different estimation methods, and it confirms consistency in findings and provides additional support for their managerial relevance.

Subsequently, we investigated the effect of SMT service climate at T1 on several team outcomes at T2. To assess the linkage between SMT service climate and its consequences empirically, we aggregated employee and customer perceptions to the group level. It has been argued that customers primarily observe the outcome of integrative working relationships among multiple employees (Allen and Grisaffe 2001). Furthermore, outgroup-homogeneity theory postulates that people tend to perceive other groups as more uniform than their own group (Quattrone and Jones 1980). This means that whereas employees may have a relatively detailed view of their own team's performance, external customers are likely to generalize the quality of service offered by one or multiple team members as a common characteristic of a homogeneous out-group. Moreover, it was not possible to match empirically employee and customer evaluations and productivity criteria at the individual level of analysis.

We tested H₆-H₈ using a multivariate regression model formulated as a two-level hierarchical linear model. Level 1 refers to the dependent variables indexed by h = 1, ..., m, and Level 2 reflects the teams j = 1, ..., N. Thus, each measurement of a dependent variable for some team is indicated by a separate line in the data matrix, which contains the values j, h, Y_hj, x_1j and those of other explanatory variables. To formulate the multivariate regression model as a hierarchical linear model, we used the dummy variables d₁ to d_m to indicate the dependent variables (i.e., perceived service quality_T2, share of customer_T2, and sales productivity_T2). Dummy variable d_h is equal to one or zero, depending on whether the data line refers to the dependent variable Y_h or to another dependent variable. With the dummy variables, the regression models for the in dependent variables can be integrated into one two-level hierarchical model with the following expression:

( 1) [Multiple line equation(s) cannot be represented in ASCII text]

We multiplied all variables (including the constant) by the dummy variables. Note that in the sums over s = 1, ..., m, only the term for s = h renders a contribution; all other terms disappear. The following equation involves a simplified representation of Equation 1:

( 2) [Multiple line equation(s) cannot be represented in ASCII text]

where Y_hj is the measurement on the hth variable for group j, SERVCLIM_T1 is the team's SMT service-climate average at T1, and NROUT reflects the type of service.

Our results in Table 7 reveal significant, positive effects of SMT service climate_T1 on perceived service quality_T2 and share of customer_T2 and a significant, negative effect on sales productivity_T2. The findings support H₆ and H₇. Furthermore, the service-type dummy "nonroutine services" is positively related to customer perceived service quality but is negatively related to share of customer and sales productivity. In addition, we tested interactions between service type and the SMT service climate. There is a positive interaction effect of nonroutine services x SMT service climate on perceived service quality, but no significant interactions appear for share of customer and sales productivity. Thus, H_8a is supported, and H_8b and H_8c are rejected.

The key objective of our study was to analyze the chain of events between perceptions of boundary-spanning service employees organized in SMTs and critical marketing performance criteria. We developed the construct of SMT service climate and empirically demonstrated that it plays a pivotal role in the prediction of SMT performance dimensions. We have empirical verification that tolerance for self-management, flexibility, and inter-and intrateam support have a direct, positive impact on individual employees' service-climate perceptions. In addition, we found that the group-level measure of intrateam support adds uniquely to the explanation of SMT service-climate perceptions. Moreover, we found that service type moderates the group-level effect of flexibility. Both effects persist over time, which nuances Mohammed, Mathieu, and Bartlett's (2002) general observation that perceptual agreement among team members with respect to interpersonally oriented behaviors is crucial to the performance of the team's core tasks. For boundary-spanning service delivery, shared perceptions of support and team flexibility are conducive to the creation of a climate related to the SMT's core task. Our results suggest that the dynamics of supportive behavior in the SMT's providing service to customers have crystallized into a coherent, collective perception and that beliefs of any team member are likely to be influenced by the attitudes and behavior of other members. This seems important because service delivery is heterogeneous across employees, and delivery of consistent service quality is a constant challenge. Further more, because nonroutine services are relatively more complex and require in-depth know-how, there is a greater need to integrate mechanisms and group-embedded expertise when addressing heterogeneous, unpredictable customer demands on the front line. The group-level interaction effect signifies that a collective understanding of one another's roles and shared beliefs about the capability to perform various roles are particularly important for service-climate perceptions in SMTs that deliver nonroutine services. This extends Druskat and Pescosolido's (2002) finding that in complex task conditions, there is a higher need for information exchange from manufacturing SMTs to the setting of nonroutine service delivery. In contrast, no additional cross-level effects (i.e., beyond individual perceptions) were reported for tolerance for self-management and interteam support. This implies that boundary-spanning service employees' perceptions of the impact of climate-defining determinants are based primarily on personal cognitions and interpretations of the team environment rather than on a property of the team.

This study extends previous research by taking into account a comprehensive set of SMTs' marketing performance measures. Our results demonstrate that SMT service-climate perceptions have a positive impact on customer perceived service quality and share of customer and a negative effect on sales productivity. We also find that the positive effect of customer perceived quality is significantly stronger for nonroutine services than for routine services. Thus, the direct relationship between organizational service climate and perceived service quality in financial services reported by Schneider, White, and Paul (1998) holds at the team level, but the strength of the relationship seems to be contingent on the service type. Particularly in the delivery of nonroutine services, which is often aimed at fulfilling specific customer needs, a climate focusing on service excellence appears to provide added value from customers' perspectives. In contrast, the relatively high within-group variability of SMT service climate for the routine service setting may have weakened the relationship between SMT service climate and perceived service quality.

Although SMT service climate contributes to a higher share of customer, it is associated with lower sales productivity levels. These results and the low correlations between the effectiveness measures may indicate what has been labeled the "performance paradox," a phenomenon that has been encountered in previous studies on firm performance (Meyer and Gupta 1995) and team performance (Spreitzer, Cohen, and Ledford 1999). An SMT climate for service does not necessarily improve all aspects of frontline service performance. This confirms Anderson, Fornell, and Rust's (1997) contention that trade-offs between different performance parameters are particularly applicable to services, and it emphasizes the necessity to set off customer parameters against productivity parameters to create an optimal balance.

Our study contributes to the theoretical literature on SMTs in four ways. First, our findings illustrate the relevance of the SMT service climate in explaining the divergent findings encountered in the SMT literature. Although service climate has been studied extensively at the organizational level, the concept has been virtually unexplored in relation to SMTs. We conducted our study in the context of financial consumer services. Further research should assess the generalizability of our findings to other service settings (e.g., technical support, business-to-business services). Although we unequivocally demonstrate the impact of service climate on important performance parameters, the body of work on this construct in the marketing and organizational behavior areas suggests that there is still much to be learned about the nature and scope of the SMT service climate as it relates to team performance. Following Schneider, Salvaggio, and Subirats (2002), we modeled within-team variability by taking the standard deviation of the SMT service climate as a measure of cohesiveness. Notably, significant, positive interaction effects of SMT service climate. SMT climate cohesiveness on customer perceived service quality and sales productivity emerged. These findings demonstrate the relevance of perceptual agreement on SMT climate as a moderator of the relationship between SMT service climate and the outcome variables. However, the effect of cohesiveness did not hold when we specified conditions for service type, possibly because of relatively strong differences in within-group variability between routine and nonroutine service delivery. Additional insight is needed into the contingencies that influence the relationship between SMT service climate and effective customer service.

Second, our research contributes to a better understanding of which factors shape a climate for service within boundary-spanning SMTs. Both team-and company-related characteristics influence team members' perceptions of the SMT service climate. Our set of predictor variables pertains predominantly to interpersonally oriented behaviors. In further research, the focus might be more on the impact of technical-administrative or task-related behavior in SMTs.

Third, our study advances the understanding of the impact of predictor-criterion relationships across levels of analysis. Although recent multilevel research has recognized the importance of comparing relationships across levels (Ostroff, Kinicki, and Clark 2002), studies have focused almost exclusively on methodological issues. Additional investigation is needed to address the underlying theoretical mechanisms that cause these differences. For example, our findings on interteam support suggest the relevance of considering distortions in employee perceptions during comparison of relationships across levels. The purpose of ratings, performance norms, and social dynamics may affect ratings of situational elements (Ostroff 1993). It makes a difference whether respondents evaluate their own team or other (rival) teams within the organization. In other words, the team members' various motives and interests influence their workplace evaluation. Thus, conceptual development is required on implicit and explicit processes among individual service employees, their colleagues, and the broader organizational context.

Fourth, our findings present a balanced perspective of service climate as a success factor related to SMTs. Although the growth and importance of SMTs has been studied, research has neglected how SMTs are related to important business-performance measures. The study empirically demonstrates that all benefits of this factor may not be realized simultaneously and that inherent trade-offs are made. The divergent effects of the SMT service climate on customers' perceived service quality and share of customer, rather than sales productivity, may motivate researchers to investigate complementary mediator variables that explain the performance paradox between customer and productivity-based measures. Future conceptual models need to include more productivity-oriented performance mediators (e.g., cost consciousness) that are related to SMTs' planning and control activities (Singh 2000).

Our findings suggest several managerial implications. Given the group-level impact of intrateam processes, competency development and measurement with regard to intrateam support and flexibility should be aligned at the team level. Team-level interventions that are effective in creating a collective sense of support in the team include group exercises aimed at consensus building on, for example, service delivery standards, joint problem identification and analysis of customer complaints, the provision of assistance to colleagues in dealing with customers, and groupware and shared databases that facilitate information exchange. In addition, performance assessments should be related to the team's collective capabilities in providing internal and external service. Cross-training programs that focus on the ability of specialist employees to promote interchangeable expertise within nonroutine SMTs may be offered. Particularly for teams delivering complex and extended services, group meetings should be devoted to increasing team members' understanding of one another's tasks through role-play exercises. Cases on handling complex customer requests may be used to demonstrate problem solving and decision-making skills.

Our findings show that it is important to create a context that is supportive to self-management on the firm's front line, one that goes beyond formally loosening the managerial reigns. Tolerance for self-management entails creating a sense of personal freedom and control with respect to, for example, meeting customer expectations (see Bowen and Lawler 1992). However, the authority to perform in the interest of the customer as a team should be accompanied by the knowledge and skills as well as the resources to analyze team performance data from a business perspective, because this is one of the essential self-managing responsibilities of the team. In addition to the dissemination of power, timely feedback and strategic information on, for example, the impact of SMT decisions on customer perceptions and company profitability should be distributed and discussed within the team. Management should play a leading role in coaching SMTs in their role as empowered customer-company interfaces. This calls for active engagement in plenary discussions on the firm's service strategy; involvement in training programs for skills in problem solving and customer complaint handling, process analysis techniques, and performance assessments; and embedding of the role of SMTs in the social structure of the organization. Support from other teams is also an important driver of service-climate perceptions. Interteam support may be encouraged through job-rotation schemes, cross-training between SMTs on customer issues, and joint customer visits. Kirkman and Rosen (2000) refer to an insurance company that uses so-called bridge teams that are responsible for facilitating the communication and cooperation between service teams. Because the impact of extrateam factors occurs predominantly at the individual level, managerial strategies need not focus exclusively on team-level interventions. Personalized training in conjunction with plenary meetings and internal media may be used to reach employees and to strengthen the impact of climate-shaping antecedents for teams of frontline service providers.

A potentially counterintuitive and important finding suggests that firms should be aware of the SMT service climate's potential to have an adverse effect on sales productivity. Thus, SMTs may need both a productivity and a service orientation if management expects them to affect different outcomes positively. If the costs of service operations need to be reduced (a likely reality today), SMTs need to be involved in developing efficiency improvements, such as designing "smarter" service-delivery routines or identifying opportunities for cross-or up-selling during service encounters. When there is a need to work simultaneously on multiple drivers of different performance parameters, the results may vary across service types. Moreover, it is unlikely that all desirable outcomes will manifest at the same time. Our findings confirm that the impact of SMT climate can be observed over an extended period of time and can provide evidence of the potential to incorporate time frames in diagnostic assessments of the effectiveness of SMTs.

The authors thank Richard Feinberg, Roberta Levy, Christine Moorman, Tor Wallin Andreassen, Pratibha Dabholkar, and the four anonymous JM reviewers for their valuable comments on previous drafts of this article.

(n1) To assess the construct validity of our focal variable, we conducted an additional CFA in which we included both SMT service climate and its conceptual counterpart "overall service climate" (Schneider, White, and Paul 1998). The results (Χ² = 236.74, 43 d.f.; goodness-of-fit index = .97; adjusted goodness-of-fit index = .95; root mean square error of approximation = .057; normed fit index = .98; and comparative fit index = .98) indicate the unidimensionality of both constructs. To determine discriminant validity, we used a chi-square difference test (1 d.f.) to test for unity between the constructs. The test showed significance at p < .05.

(n2) To control for multicollinearity, we inspected the variance inflation factors of the variables. The control variables and the antecedents yielded values less than 3.9 and 2.0, respectively, indicating the absence of serious multicollinearity problems (Kleinbaum, Kupper, and Muller 1988).

(n3) Within-group deviation is similar to group mean-centering. In all analyses, we group mean-centered individual-level variables and grand mean-centered group-level variables to (1) distinguish within-and between-group variance, (2) reduce multicollinearity, and (3) facilitate model estimation (see Bryk and Raudenbush 1992).

(n4) Gibbs sampling works by simulating a new value for each parameter from its conditional distribution and by assuming that the current values for the other parameters are the true values.

Legend for Chart:

B - Employees of SMTs in Nonroutine Services (N = 316) Frequency
C - Employees of SMTs in Nonroutine Services (N = 316) Percentage
D - Employees of SMTs in Routine Services (N = 509) Frequency
E - Employees of SMTs in Routine Services (N = 509) Percentage

    A                              B      C       D      E

Sex

Male                              187    59.2     86    16.9
Female                            129    40.8    423    83.1

Age (Years)

Younger than 40                   149    47.2    303    59.5
Older than 40                     167    52.8    206    40.5

Education Level

Secondary                          59    18.7    131    25.7
Tertiary                          257    81.3    378    74.3

Organizational Tenure (Years)

Less than 5                       123    38.9    242    47.5
More than 5                       193    61.1    267    52.5

Team Experience (Years)

Less than 2                       146    46.2    166    32.6
More than 2                       170    53.8    343    67.4

Job

Full time                         217    68.7    217    42.6
Part time                          99    31.3    292    57.4

Legend for Chart:

B - SMT Characteristics Nonroutine Services
C - SMT Characteristics Routine Services

           A                    B

                                C

Team size (mean [s.d.])   13.98 (5.85)

                          36.76 (12.70)

Functional areas          General financial advisers
                          Mortgage advisers
                          Portfolio advisers
                          Financial planners
                          Insurance advisers
                          Clerical staff
                          Risk analysts

                          Client services employees
                          Cashiers
                          Receptionists
                          Travel services
                          Clerical staff
                          Customer services (customer
                          complaint management)

Responsibilities          Division and allocation of work
                          Planning and budgeting
                          Recruitment and training
                          Developing new work routines
                          Monitoring member and team performance
                          Service recovery
                          Internal and external coordination

                          Division and allocation of work
                          Planning and budgeting
                          Recruitment and training
                          Developing new work routines
                          Monitoring member and team performance
                          Service recovery
                          Internal and external coordination

Notes: In essence, the responsibilities of the SMTs do not differ
between routine and nonroutine services.

Legend for Chart:

A - Measures
C - Item Mean
D - Loading
E - t-Value

     A                                        C       D       E

Employee Data(a)

SMT Service Climate (n = 6; α =
.86 at T1; α = .88 at T2)

1. Our team is continually working to        5.81    .78    25.54
improve the quality of service we
provide to our customers.

2. Our team has specific ideas about         5.32    .67    20.94
how to improve the quality of service
we provide to customers.

3. Our team often makes suggestions          5.01    .51    14.85
about how to improve the service quality
of our organization.

4. In our team we put a lot of effort        5.68    .83    28.22
in attempting to satisfy customer
expectations.

5. No matter how we feel, we always          5.45    .71    22.55
put ourselves out for every customer
we serve.

6. Within our team, employees often          5.30    .73    23.23
go out of their way to help customers.

Tolerance for Self-Management (n = 6;
α = .89 at T1; α = .90 at
T2)

1. In our team we are permitted to use       5.27    .71    22.50
our own judgment in solving problems.

2. In our team we are encouraged to          6.09    .64    19.48
take initiative.

3. Our team is allowed a high degree         5.83    .81    27.28
of initiative.

4. In our team we are allowed complete       5.54    .86    29.94
freedom in our work.

5. In our team we are allowed to do          5.25    .76    24.73
our work the way we think best.

6. As a team we are able to handle all       5.22    .72    23.09
tasks assigned to us.

Flexibility of Team Members (n = 4;
α = .68 at T1; α = .74 at
T2)

1. In our team it is easy to stand in        5.47    .66    18.03
for each other.

2. Most team members know each other's       5.97    .58    15.49
tasks.

3. I have much confidence that my team       5.69    .66    17.94
members would be able to take over my
activities.

4. Exchanging team roles and                 4.68    .48    12.50
responsibilities causes few problems.

Interteam Support (n = 7; α =
.83 at T1; α = .83 at T2)(One
Reversed Item)

1. Other teams act in a responsive           4.44    .55    16.03
manner when we forward customer
complaints.

2. The knowledge of other teams assists      4.93    .56    16.56
us in serving customers.

3. The quality of service delivered          4.64    .75    23.75
by other teams to our team is good.

4. Because of insufficient feedback          3.46    .53    15.28
from other teams our service to customers
is substandard. (reversed item)

5. Other teams provide good feedback         4.66    .54    15.80
on how to serve customers.

6. The cooperation between teams within      4.32    .85    28.56
the bank is good.

7. The employees of other teams are          4.71    .75    24.09
helpful in solving problems of customers.

Intrateam Support (n = 4; α =
.66 at T1; α = .68 at T2)

1. In our team we help each other in         5.06    .68    19.88
serving the customer.

2. The mutual support of team members        5.18    .79    23.57
is highly valued.

3. Each team member is personally            4.98    .51    15.03
responsible for the assistance of other
members in serving the customer.

4. In our team members need not formally     5.04    .52    14.67
be monitored with regard to the
assistance of colleagues.

Customer Data

Legend for Chart:

A - Measures
B - Item Mean
C - Loading
D - t-Value

         A                                      B      C      D

Customer Perceived Service Quality (n = 8;
α = .92)

1. The extent to which employees make clear
appointments                                   4.10   .62   26.17

2. Speed at which the promised information
is provided                                    4.04   .65   27.46

3. The friendliness and politeness of
employees                                      4.38   .72   31.71

4. The competence of the service by employees  4.07   .82   37.20

5. The time taken by employees to serve you    4.34   .75   33.12

6. The attention employees pay to you          4.25   .82   37.79

7. The extent to which employees show empathy  4.03   .81   37.83

8. The readiness of the employees to help you  4.25   .86   40.41

Legend for Chart:

A - Share of Customer
B - Nonroutine Services
C - Routine Services

        A                             B

                                      C

Average customer usage       Mortgages, loans (53.6%)
rates (%) of the different   Investment funds (52.7%)
services offered             Stocks (43.9%)
                             Insurance (46.5%)

                             Checking account (95.2%)
                             Savings account (79.1%)
                             Electronic/telephone banking (30.3%)
                             Currency exchange (20.1%)
                             Credit application accounts (44.9%)
                             Travel services (18.7%)

(a) The CFA is based on employee data collected at T1.

Notes: All t-values are significant at p<.05. For employee data,
fit indexes are χ² = 925.40, 314 d.f.; goodness-of-fit
index (GFI) = .92; adjusted goodness-of-fit index (AGFI) = .91;
root mean square error of approximation (RMSEA) = .049; normed
fit index (NFI) = .90; and comparative fit index (CFI) = .93.
For customer data, fit indexes are χ² = 44.64, 20 d.f.;
GFI = .99; AGFI = .98; RMSEA = .036; NFI = .99; and CFI = 1.00.

Legend for Chart:

A - Variables
B - Mean (s.d.)
C - 1
D - 2
E - 3
F - 4
G - 5
H - 6
I - 7
J - 8
K - 9
L - 10
M - 11

  A
                   B                C           D           E
                                    F           G           H
                                    I           J           K
                                                L           M

1. Education
              4.23  (1.60)(a)     --        -.07         .01
                                -.57(***)    .35(***)    .68(***)
                                 .19         .42(***)   -.06
                                            -.07         .01

2. Tenure
              3.41  (1.67)(b)   -.28(***)     --         .71(***)
                                -.18         .19         .22(*)
                                -.08         .01         .00
                                            -.21        -.10

3. Age
              2.62   (.85)(c)   -.16(***)    .70(***)     --
                                -.15         .24(*)      .27(**)
                                 .18         .08         .15
                                            -.31(**)    -.01

4. Team size
             20.44 (13.54)(d)     --          --          --
                                  --        -.40(***)   -.69(***)
                                -.06        -.34(***)    .09
                                            -.05         .08

5. Front office (%)
             47.03 (14.21)(d)     --          --          --
                                  --          --         .46(***)
                                 .19         .27(**)    -.04
                                            -.16         .01

6. Nonroutine services
               .59   (.50)(d)     --          --          --
                                  --          --          --
                                 .13         .34(***)   -.04
                                            -.15        -.10

7. SMT service climate
              5.43   (.82)       .02        -.00         .02
                                  --          --          --
                                  --         .56(***)    .48(***)
                                             .08         .62(***)

8. Tolerance for self-management
              5.54   (.89)       .09(**)     .04         .02
                                  --          --          --
                                 .37(***)     --         .25(*)
                                             .10         .34(***)

9. Flexibility of members
              5.45   (.89)      -.00        -.01         .01
                                  --          --          --
                                 .38(***)    .24(***)     --
                                             .18         .53(***)

10. Interteam support
              4.45   (.93)      -.07(**)    -.07(**)    -.10(***)
                                  --          --          --
                                 .28(***)    .17(***)    .26(***)
                                              --         .33(**)

11. Intrateam support
              5.07   (.94)      -.03         .03         .06
                                  --          --          --
                                 .49(***)    .32(***)    .46(***)
                                             .33(***)     --

(*) p < .10.
(**) p < .05.
(***) p < .001.

(a) Education consists of seven categories that include various
forms of secondary and tertiary education.

(b) Tenure consists of six categories that range from "<1 year"
to ">5 years."

(c) Age consists of six categories that range from "<21 years
old" to ">60 years old."

(d) Means (s.d.) of team size, front-office work, and nonroutine
services are based on the group averages.

Notes: N = 825 respondents of 61 groups. Individual-level
correlations are in the lower triangle, and group-level
correlations are in the upper triangle. Correlations in the
upper triangle are the correlations between the group averages.

Legend for Chart:

A - Variables
B - Mean (s.d.)
C - N = 61 1
D - N = 61 2
E - N = 61 3
F - N = 61 4
G - N = 61 5
H - N = 61 6
I - N = 61 7
J - N = 61 8
K - N = 61 9
L - N = 61 10
M - N = 61 11
N - N = 61 12
O - N = 61 13
P - N = 61 14

 A
             B              C              D            E
                            F              G            H
                            I              J            K
                            L              M            N
                                           O            P

1. Perceived service quality<sub>T2</sub>
       4.17    (.16)       --
                                                      .20
                          .11          -.02           .07
                                       -.07(i)        .14(i)
                                       -.13(i)       -.05(i)

2. Share of customer<sub>T2</sub>
       2.34    (.65)(a)  -.14            --
                                                     -.23(*)
                          .08           .33(**)       .30(**)
                                       -.14(i)       -.14(i)
                                        .14(i)        .19(i)

3. Sales productivity<sub>T2</sub>
     525.73 (477.92)(a)  -.25(*)        .63(***)       --
                                                     -.33(**)
                          .07           .13           .19
                                       -.26(*)(i)    -.05(i)
                                        .11(i)        .27(**)(i)

4. Nonroutine services
        .59    (.50)      .24(*)       -.69(***)     -.97(***)
                           --

5. SMT service climate<sub>T1</sub>
       5.42    (.35)      .40(***)      .12          -.21
                          .13            --

6. Tolerance for self-management<sub>T1</sub>
       5.52    (.43)      .37(***)     -.12          -.38(***)
                          .34(***)      .56(***)       --

7. Flexibility of members<sub>T1</sub>
       5.41    (.40)      .28(**)       .13          -.04
                         -.04           .48(***)      .25(*)
                           --

8. Interteam support<sub>T1</sub>
       4.44    (.48)      .02           .33(***)      .11
                         -.15           .08           .10
                          .18            --

9. Intrateam support<sub>T1</sub>
       5.03    (.38)      .30(**)       .31(**)       .02
                         -.10           .62(***)      .34(***)
                          .53(***)      .33(**)        --

10. SMT service climate<sub>T2</sub>
       5.44    (.35)      .25(*)(i)    -.11(i)       -.32(**)(i)
                          .29(**)(i)    .63(***)(i)   .50(***)(i)
                          .42(***)(i)   .07(i)        .36(***)(i)
                           --

11. Tolerance for self-management<sub>T2</sub>
       5.61    (.41)      .06(i)       -.18(i)       -.37(***)(i)
                          .34(**)(i)    .32(**)(i)    .62(***)(i)
                          .16(i)        .06(i)        .14(i)
                          .47(***)(i)    --

12. Flexibility of members<sub>T2</sub>
       5.51    (.33)      .23(*)(i)    -.18(i)       -.19(i)
                          .16(i)        .42(***)(i)   .36(***)(i)
                          .38(***)(i)  -.08(i)        .30(**)(i)
                          .46(***)(i)   .40(***)(i)    --

13. Interteam support<sub>T2</sub>
       4.54    (.50)     -.08(i)        .12(i)        .04(i)
                         -.04(i)       -.06(i)        .03(i)
                          .19(i)        .57(***)(i)   .20(i)
                          .19(i)        .39(***)(i)   .32(**)(i)
                                         --

14. Intrateam support<sub>T2</sub>
       5.17    (.52)      .06(i)        .12(i)        .10(i)
                         -.11(i)        .40(***)(i)   .14(i)
                          .11(i)        .24(i)        .36(***)(i)
                          .41(***)(i)   .32(**)(i)    .50(***)(i)
                                        .40(***)(i)    --

(*) p < .10.
(**) p < .05.
(***) p < .001.

(a) There are large differences between routine and nonroutine
services with respect to share of customer (routine services, mean
[s.d] = 1.97 [.43]; nonroutine services, mean [s.d.] = 2.88 [.54])
and sales productivity (routine services, mean [s.d] = 1076.26
[171.32]; nonroutine services, mean [s.d] = 143.37 [67.77]).

Notes: (i) For italicized correlations, N = 56. Correlations
among variables are represented in the lower triangle.
Coefficients in the upper triangle are the partial correlations
between antecedents and outcomes. For the partial correlations
with antecedents at T1, we partialled out the effect of SMT
service climate<sub>T1</sub>; for antecedents at T2, we
partialled out the effect of SMT service climate<sub>T2</sub>.

Legend for Chart:

B - Model A1 SMT Service Climate<sub>T1</sub> Coefficient
    (Standard Errors)(a)
C - Model A1 SMT Service Climate<sub>T1</sub> Magnitude
    Difference(b)
D - Model A2 SMT Service Climate<sub>T1</sub> Coefficient
    (Standard Errors)(a)
E - Model A2 SMT Service Climate<sub>T1</sub> Magnitude
    Difference(b)
F - Hypothesis

   A
                 B                   C                  F
                 D                   E

Intercept
            .573 (.713)
            .827 (.766)

Individual-Level Control Variables

Education
           -.017 (.017)
           -.018 (.017)

Tenure
           -.027 (.021)
           -.021 (.021)

Age
            .040 (.039)
            .036 (.040)

Group-Level Control Variables

Education
           -.015 (.063)         .003 (.065)
           -.011 (.067)         .006 (.070)

Tenure
           -.225 (.082)(**)    -.198 (.085)(**)
           -.219 (.084)(**)    -.198 (.087)(*)

Age
            .361 (.153)(**)     .322 (.159)(*)
            .317 (.160)(*)      .277 (.167)(*)

Team size
            .004 (.003)
            .004 (.003)

Front office (%)
           -.082 (.199)
           -.085 (.208)

Nonroutine services
            .179 (.101)(*)
            .182 (.106)(*)

Individual-Level Antecedents

Tolerance for self-management
            .220 (.032)(**)                        H<sub>1</sub>
            .224 (.031)(**)

Flexibility of team members
            .163 (.033)(**)                        H<sub>2</sub>
            .170 (.033)(**)

Interteam support
            .133 (.034)(**)                        H<sub>3a</sub>
            .134 (.034)(**)

Intrateam support
            .238 (.031)(**)                        H<sub>3b</sub>
            .231 (.032)(**)

Group-Level Antecedents

Tolerance for self-management
            .219 (.080)(**)     .000 (.086)        H<sub>4a</sub>
            .193 (.082)(**)    -.021 (.088)

Flexibility of team members
            .160 (.094)(*)     -.008 (.099)        H<sub>4b</sub>
            .217 (.099)(*)      .055 (.105)

Interteam support
           -.006 (.071)        -.141 (.077)(*)     H<sub>4c</sub>
           -.039 (.074)        -.172 (.080)(*)

Intrateam support
            .497 (.100)(**)     .262 (.105)(**)    H<sub>4d</sub>
            .460 (.102)(**)     .224 (.107)(*)

Cross-Level Interactions: Individual-Level
Antecedents x Nonroutine Service Type

Tolerance
           -.035 (.025)

Flexibility
           -.042 (.028)

Interteam
           -.021 (.027)

Intrateam
            .034 (.027)

Group-Level Interactions: Group-Level
Antecedents x Nonroutine Service Type

Tolerance
                                                   H<sub>5a</sub>
           -.007 (.031)        -.003 (.031)


Flexibility
                                                   H<sub>5b</sub>
            .074 (.033)(*)      .086 (.034)(**)

Interteam
                                                   H<sub>5c</sub>
           -.002 (.031)         .002 (.031)

Intrateam
                                                   H<sub>5d</sub>
           -.047 (.036)        -.054 (.037)

Increase in Model Fit

           χ² = 362.72 (21 d.f.)(**)

           χ² = 11.41 (8 d.f.)

Explained Variance (%)

Individual level
                       37.9%
                       38.2%

Group level
                       58.9%
                       60.0%

(*) p < .05.
(**) p < .01.

(a) Unstandardized regression coefficients.

(b) We tested differences in magnitude between individual- and
group-level coefficients by means of raw-score analyses, and
these are reflected in the presented group-level coefficients.

Notes: Significance is based on one-tailed tests.

Legend for Chart:

B - Model B1 (Lagged Analysis) SMT Service Climate<sub>T1</sub>
    Coefficient (Standard Errors)(a)
C - Model B1 (Lagged Analysis) SMT Service Climate<sub>T1</sub>
    Magnitude Difference(b)
D - Model B2 (Gibbs Sampling)(c) SMT Service Climate<sub>T2</sub>
    Coefficient (Standard Errors)(a)
E - Model B2 (Gibbs Sampling)(c) SMT Service Climate<sub>T2</sub>
    Magnitude Difference(b)

      A
                                B                    C
                                D                    E

Intercept
                          -.255 (.967)
                           .842 (.773)

SMT Service Climate<sub>T1</sub>
                           .334 (.051)(**)

Individual-Level Control Variables

Education
                          -.001 (.024)
                          -.018 (.017)

Tenure
                           .000 (.030)
                          -.021 (.021)

Age
                           .141 (.060)(**)
                           .035 (.040)

Group-Level Control Variables

Education
                          -.064 (.086)         -.045 (.094)
                          -.012 (.069)          .006 (.071)

Tenure
                          -.189 (.092)(*)      -.192 (.102)(*)
                          -.219 (.085)(**)     -.196 (.088)(*)

Age
                           .496 (.188)(**)      .363 (.207)(*)
                           .316 (.163)(*)       .278 (.169)(*)

Team size
                           .004 (.004)
                           .004 (.003)

Front office (%)
                          -.308 (.280)
                          -.088 (.214)

Nonroutine services
                           .265 (.145)(*)
                           .182 (.107)(*)

Individual-Level Antecedents

Tolerance for self-management
                           .138 (.058)(**)
                           .223 (.032)(**)

Flexibility of team members
                           .164 (.073)(*)
                           .169 (.034)(**)

Interteam support
                           .089 (.049)(*)
                           .132 (.031)(**)

Intrateam support
                           .100 (.051)(*)
                           .232 (.031)(**)

Group-Level Antecedents

Tolerance for self-management
                           .183 (.099)(*)       .019 (.118)
                           .194 (.084)(*)       -.021 (.090)

Flexibility of team members
                           .121 (.119)          .040 (.136)
                           .214 (.101)(*)       .064 (.106)

Interteam support
                           .009 (.087)         -.076 (.106)
                          -.039 (.075)         -.168 (.083)(*)

Intrateam support
                           .335 (.121)(**)      .242 (.138)(*)
                           .459 (.105)(**)      .214 (.109)(*)

Cross-Level Interactions: Individual-Level
Antecedents x Nonroutine Service Type

Tolerance
                          -.009 (.038)
                          -.035 (.025)

Flexibility
                           .052 (.053)
                          -.041 (.027)

Interteam
                           .001 (.037)
                          -.023 (.025)

Intrateam
                          -.054 (.039)
                           .035 (.027)

Group-Level Interactions: Group-Level
Antecedents x Nonroutine Service Type

Tolerance
                          -.014 (.046)         -.003 (.050)
                          -.007 (.031)         -.002 (.032)

Flexibility
                           .078 (.044)(*)       .083 (.048)(*)
                           .074 (.033)(*)       .082 (.034)(**)

Interteam
                           .066 (.038)(*)       .060 (.041)
                          -.002 (.032)          .002 (.032)

Intrateam
                          -.082 (.052)         -.058 (.057)
                          -.046 (.037)         -.054 (.038)

Explained Variance (%)

Individual level
                           42.3%
                           37.7%

Group level
                           59.0%
                           57.9%

(*) p < .05.
(**) p < .01.

(a) Unstandardized regression coefficients.

(b) We tested differences in magnitude between individual- and
group-level coefficients by means of raw-score analyses, and
these are reflected in the presented group-level coefficients.

(c) Model B2: Gibbs sampling (10,000 iterations).

Notes: Significance is based on one-tailed tests.

Legend for Chart:

B - Model C1 Coefficient (Standard Errors)(a)
C - Model C2 Coefficient (Standard Errors)(a)
D - Hypothesis

     A

                                 B
                                 C                     D

SERVCLIMT1 → QUAL<sub>T2</sub>(b)

                           .941     (.308)(**)
                           .969     (.295)(**)    H<sub>6a</sub>

NROUT → QUAL<sub>T2</sub>

                           .501     (.313)
                           .510     (.299) (*)

SERVCLIM<sub>T1</sub> x NROUT → QUAL<sub>T2</sub>

                           .336     (.134)(**)    H<sub>8a</sub>

SERVCLIM<sub>T1</sub> → SHARE<sub>T2</sub>

                           .275     (.119)(**)
                           .268     (.118)(**)    H<sub>6b</sub>

NROUT → SHARE<sub>T2</sub>

                          -.953     (.121)(**)
                          -.955     (.119)(**)

SERVCLIM<sub>T1</sub> x NROUT → SHARE<sub>T2</sub>

                          -.082     (.053)        H<sub>8b</sub>

SERVCLIM<sub>T1</sub> → SPROD<sub>T2</sub>

                        -73.862   (30.076)(**)
                        -72.430   (29.961)(**)    H<sub>7</sub>

NROUT → SPROD<sub>T2</sub>

                       -922.734   (30.539)(**)
                       -922.305   (30.403)(**)

SERVCLIM<sub>T1</sub> x NROUT → SPROD<sub>T2</sub>

                         16.811   (13.570)        H<sub>8c</sub>

Residual Between-Groups
Covariance Terms

σ²<sub>h</sub> = var(e<sub>hj</sub>), (h = 1)

                          1.417     (.256)
                          1.298     (.235)

σ²<sub>h</sub> = var(e<sub>hj</sub>), (h = 2)

                           .211     (.039)
                           .206     (.037)

σ²<sub>h</sub> = var(e<sub>hj</sub>), (h = 3)

                     13,507.870 (2440.253)
                     13,385.760 (2423.142)

σ<sub>12</sub> = cov(e<sub>1j</sub>, e<sub>2j</sub>)

                          -.041     (.070)
                          -.006     (.066)

σ<sub>13</sub> = cov(e<sub>1j</sub>, e<sub>3j</sub>)

                          6.225   (17.704)
                          -.871   (16.872)

σ<sub>23</sub> = cov(e<sub>2j</sub>, e<sub>3j</sub>)

                         -7.755    (6.927)
                         -6.144    (6.773)

Increase in model fit:(c)

                   χ² = 232.45(6 d.f.)(**)

                   χ² = 10.32 (3 d.f.)(*)

Increase in model fit:(d)

                      χ² = 1.71(3 d.f.)

                      χ² = .85 (3 d.f.)

(*) p < .05.
(**) p < .01.

(a) Unstandardized regression coefficients.

(b) As we aggregated the scale of perceived service quality to
the group level, we calculated r<sub>WG(J)</sub>-, ICC
(1)-, and ICC (2)- coefficients (respectively .96, .04, and .51)
to determine within-group agreement and interdependence.

(c) Increase in model fit with inclusion of the predictor
variables.

(d) Increase in model fit with inclusion of the covariance terms
among the outcome variables.

Notes: Significance is based on one-tailed tests.

DIAGRAM: FIGURE 1; Conceptual Framework

Allen, Nathalie J. and Douglas B. Grisaffe (2001), "Employee Commitment to the Organization and Customer Reactions: Mapping the Linkages," Human Resource Management Review , 11 (3), 209-236.

Anderson, Eugene W., Claes Fornell, and Roland T. Rust (1997), "Customer Satisfaction, Productivity, and Profitability: Differences Between Goods and Services," Marketing Science , 16 (2), 129-45.

Babin, Barry J. and Jill A. Attaway (2000), "Atmospheric Affect as a Tool for Creating Value and Gaining Share of Customer," Journal of Business Research , 49 (2), 91-99.

Balkema, Ageeth and Eric Molleman (1999), "Barriers to the Development of Self-Organizing Teams," Journal of Managerial Psychology , 14 (1-2), 230-43.

Baron, Reuben M. and David A. Kenny (1986), "The Moderator-Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations," Journal of Personality and Social Psychology , 51 (6), 1173-82.

Batt, Rosemary (1999), "Work Organization, Technology, and Performance in Customer Service and Sales," Industrial and Labor Relations Review , 52 (4), 539-63.

----- (2002), "Managing Customer Services: Human Resource Practices, Quit Rates, and Sales Growth," Academy of Management Journal , 45 (3), 587-97.

Bernhardt, Kenneth L., Naveen Donthu, and Pamela A. Kennett (2000), "A Longitudinal Analysis of Satisfaction and Profitability," Journal of Business Research , 47 (2), 161-71.

Berry, Leonard L., A. Parasuraman, and Valarie Zeithaml (1994), "Improving Service Quality in America: Lessons Learned," Academy of Management Executive , 8 (2), 32-45.

Bettencourt, Lance A. and Stephen W. Brown (1997), "Contact Employees: Relationships Among Workplace Fairness, Job Satisfaction, and Prosocial Service Behaviors," Journal of Retailing, 73 (1), 39-61.

Blau, Gary (1995), "Influence of Group Lateness on Individual Lateness: A Cross-Level Examination," Academy of Management Journal , 38 (5), 1483-96.

Bliese, Paul D. (2000), "Within-Group Agreement, Non-Independence, and Reliability: Implications for Data Aggregation and Analysis," in Multilevel Theory, Research, and Methods in Organizations , Katherine J. Klein and Stephen W.J. Kozlowsik, eds. San Francisco: Jossey-Bass, 349-81.

Bolton, Ruth N. and James H. Drew (1991), "A Longitudinal Analysis of the Impact of Service Changes on Customer Attitudes," Journal of Marketing , 55 (January), 1-9.

Bowen, David E. and Edward E. Lawler III (1992), "The Empowerment of Service Workers: What, Why, How, and When?" Sloan Management Review , 33 (3), 31-39.

Brislin, Richard W. (1980), "Translation and Content Analysis of Oral and Written Materials," in Handbook of Cross-Cultural Psychology , Vol. 2, Harry C. Triandis and John W. Berry, eds. Boston: Allyn & Bacon, 389-444.

Brown, Steven P. and Thomas W. Leigh (1996), "A New Look at Psychological Climate and Its Relationship to Job Involvement, Effort, and Performance," Journal of Applied Psychology , 81 (4), 358-68.

Bryk, Anthony S. and Stephen W. Raudenbush (1992), Hierarchical Linear Models: Applications and Data Analysis Methods. Newbury Park, CA: Sage Publications.

Cameron, Marsha and Jack Boise (1995), "Linking Rewards to Service: The Welch's Case," Journal of Compensation & Benefits, 11 (2), 60-64.

Campion, Michael A., Gina J. Medsker, and Catherine A. Higgs (1993), "Relations Between Work Group Characteristics and Effectiveness: Implications for Designing Effective Work Teams," Personnel Psychology , 46 (4), 823-50.

Chan, David (1998), "Functional Relations Among Constructs in the Same Content Domain at Different Levels of Analysis: A Typology of Composition Models," Journal of Applied Psychology, 83 (2), 234-46.

Chaston, Ian (1998), "Self-Managed Teams: Assessing the Benefits for Small Service-Sector Firms," British Journal of Management, 9 (1), 1-12.

Cohen, Susan G., Lei Chang, and Gerald E. Ledford Jr. (1997), "A Hierarchical Construct of Self-Management Leadership and Its Relationship to Quality of Work Life and Perceived Work Group Effectiveness," Personnel Psychology , 50 (2), 275-308.

Cook, Justine D., Stephen J. Hepworth, Toby D. Wall, and Peter B. Warr (1981), The Experience of Work: A Compendium and Review of 249 Measures and Their Use . London: Academic Press.

Davis, Tim R.V. (1999), "Different Service Firms, Different Core Competencies," Business Horizons , 42 (5), 23-34.

Druskat, Vanessa U. and Anthony T. Pescosolido (2002), "The Content of Effective Teamwork Mental Models in Self-Managing Teams: Ownership, Learning, and Heedful Interrelating," Human Relations , 55 (3), 283-314.

Gilks, Walter, Sylvia Richardson, and David J. Spiegelhalter (1996), Markov Chain Monte Carlo in Practice . London: Chapman and Hall.

Gilson, Lucy L., Christina E. Shalley, and Terry C. Blum (2001), "Team and Organizational Attitudes as a Lens and Mirror Impacting Customer Satisfaction: An Empirical Test in Self-Managed Teams," Journal of Quality Management , 6 (2), 235-56.

Griffin, Ricky (1991), "Effects of Work Redesign on Employee Perceptions, Attitudes, and Behaviors: A Long-Term Investigation," Academy of Management Journal , 32 (2), 425-35.

Gully, Stanley M., Aparna Joshi, Kara A. Incalcaterra, and J. Mattew Beaubien (2002), "A Meta-Analysis of Team-Efficacy, Potency, and Performance: Interdependence and Level of Analysis as Moderators of Observed Relationships," Journal of Applied Psychology , 87 (5), 819-32.

Hackman, J. Richard (1987), "The Design of Work Teams," in Handbook of Organizational Behavior , Jay W. Lorsch, ed. Englewood Cliffs, NJ: Prentice Hall, 315-42.

----- (1992), "Group Influences on Individuals in Organizations," in Handbook of Industrial & Organizational Psychology, Marvin D. Dunnette and Leaetta M. Hough, eds. Palo Alto, CA: Consulting Psychologists Press, 199-267.

Hartline, Michael D. and O.C. Ferrell (1996), "The Management of Customer-Contact Service Employees: An Empirical Investigation," Journal of Marketing , 60 (October), 52-70.

Hinkle, Dennis E., William Wiersma, and Stephen. G. Jurs (1994), Applied Statistics for the Behavioral Sciences . Boston: Houghton Mifflin.

Horwitz, Frank and Mark A. Neville (1996), "Organization Design for Service Excellence: A Review of the Literature," Human Resource Management , 35 (4), 471-92.

James, Lawrence R. (1982), "Aggregation Bias in Estimates of Perceptual Agreement," Journal of Applied Psychology , 67 (2), 219-29.

James, Lois A. and Lawrence R. James (1989), "Integrating Work Environment Perceptions: Explorations into the Measurement of Meaning," Journal of Applied Psychology , 74 (5), 739-51.

Katz, Daniel and Robert L. Kahn (1978), The Social Psychology of Organizations . New York: John Wiley & Sons.

Kirkman, Bradley L. and Benson Rosen (2000), "Powering up Teams," Organizational Dynamics , 28 (3), 48-65.

Kleinbaum, David G., Lawrence L. Kupper, and Keith E. Muller (1988), Applied Regression Analysis and Other Multivariable Methods . Boston: PWS-Kent.

Liang, Diane, Dick W. Moreland, and Linda Argote (1995), "Group Vs. Individual Training and Group Performance: The Mediating Role of Transactive Memory," Personality and Social Psychology Bulletin , 21 (4), 384-93.

Lindell, Michael K. and Christina J. Brandt (2000), "Climate Quality and Climate Consensus as Mediators of the Relationship Between Organizational Antecedents and Outcomes," Journal of Applied Psychology , 85 (3), 331-48.

Marks, Michelle A., C. Shawn Burke, Mark J. Sabella, and Stephen J. Zaccaro (2002), "The Impact of Cross-Training on Team Effectiveness," Journal of Applied Psychology , 87 (1), 3-13.

-----, John E. Mathieu, and Stephen J. Zaccaro (2001), "A Conceptual Framework and Taxonomy of Team Processes," Academy of Management Review , 26 (3), 356-76.

Mathieu, John E., Tonia S. Heffner, Gerald F. Goodwin, Eduardo Salas, and Janis A. Cannon-Bowers (2000), "The Influence of Shared Mental Models on Team Process and Effectiveness," Journal of Applied Psychology , 85 (2), 273-83.

----- and Stacey Kohler (1990), "A Cross-Level Examination of Group Absence Influence on Individual Absence," Journal of Applied Psychology , 75 (2), 217-20.

Meyer, Marshall W. and Vipin Gupta (1995), "The Performance Paradox," Research in Organizational Behavior , 16 (3), 309-369.

Mohammed, Susan, John E. Mathieu, and A.L. "Bart" Bartlett (2002), "Technical-Administrative Task Performance, Leadership Performance, and Contextual Performance: Considering the Influence of Team-and Task-Related Composition Variables," Journal of Organizational Behavior , 23 (7), 795-814.

Nunnally, Jum C. and Ira H. Bernstein (1994), Psychometric Theory, 3d ed. New York: McGraw-Hill

Ostroff, Cheri (1993), "Rater Perceptions, Satisfaction and Performance Ratings," Journal of Occupational and Organizational Psychology , 66 (4), 345-56.

-----, Angelo J. Kinicki, and Mark A. Clark (2002), "Substantive and Operational Issues of Response Bias Across Levels of Analysis: An Example of Climate-Satisfaction Relationship," Journal of Applied Psychology , 87 (2), 355-68.

Parasuraman, A., Valarie A. Zeithaml, and Leonard L. Berry (1988), "SERVQUAL: A Multiple-Item Scale for Measuring Consumer Perceptions of Service Quality," Journal of Retailing, 64 (1), 12-40.

Peccei, Riccardo and Patrice Rosenthal (1997), "The Antecedents of Employee Commitment to Customer Service: Evidence from a UK Service Context," The International Journal of Human Resource Management , 8 (1), 66-86.

Poole, Marshall S. and Robert D. McPhee (1983), "A Structurational Analysis of Organizational Climate," in Communication and Organizations: An Interpretive Approach , Linda L. Putnam and Michael E. Pacanowsky, eds. Beverly Hills, CA: Sage Publications, 195-219.

Quattrone, George A. and Ervin E. Jones (1980), "The Perception of Variability Within In-Groups and Out-Groups: Implications for the Law of Small Numbers," Journal of Personality and Social Psychology , 38 (1), 141-52.

Rasbash, Jon, William Browne, Harvey Goldstein, Min Yang, Ian Plewis, Michael Healy, Geoff Woodhouse, David Draper, Ian Langford, and Toby Lewis (2000), A User's Guide to MlwiN. London: Multilevel Models Project Institute of Education, University of London.

Schneider, Benjamin, Steven D. Ashworth, Andrew Higgs, and Linda Carr (1996), "Design, Validity, and Use of Strategically Focused Employee Attitude Surveys," Personnel Psychology , 49 (3), 695-705.

----- and David E. Bowen (1985), "Employee and Customer Perceptions of Service in Banks: Replication and Extension," Journal of Applied Psychology , 70 (3), 423-33.

-----, Amy Nicole Salvaggio, and Montse Subirats (2002), "Climate Strength: A New Direction for Climate Research," Journal of Applied Psychology , 87 (2), 220-29.

-----, Jill K. Wheeler, and Jonathan F. Cox (1992), "A Passion for Service: Using Content Analysis to Explicate Service Climate Themes," Journal of Applied Psychology , 77 (5), 705-716.

-----, Susan S. White, and Michelle C. Paul (1998), "Linking Service Climate and Customer Perceptions to Service Quality: Test of a Causal Model," Journal of Applied Psychology , 83 (2), 150-63.

Singh, Jagdip (2000), "Performance Productivity and Quality of Frontline Employees in Service Organizations," Journal of Marketing , 64 (April), 15-34.

Snijders, Tom A.B. and Roel J. Bosker (1999), Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling . London: Sage Publications.

Soteriou, Andreas and Stavros A. Zenios (1999), "Operations, Quality, and Profitability in the Provision of Banking Services," Management Science , 45 (9), 1221-38.

Spreitzer, Gretchen M., Susan G. Cohen, and Gerald E. Ledford Jr. (1999), "Developing Self-Managing Work Teams in Service Organizations," Group & Organization Management , 24 (3), 340-66.

Stewart, Greg L. and Murray R. Barrick (2000), "Team Structure and Performance: Assessing the Mediating Role of Intrateam Process and the Moderating Role of Task Type," Academy of Management Journal , 43 (2), 135-48.

Sundstrom, Eric, Kenneth P. de Meuse, and David Futrell (1990), "Work Teams: Applications and Effectiveness," American Psychologist, 45 (2), 120-33.

Tesluk, Paul E., James L. Farr, John E. Mathieu, and Robert J. Vance (1995), "Generalization of Employee Involvement Training to the Job Setting: Individual and Situational Effects," Personnel Psychology , 48 (3), 607-632.

Van Mierlo, Heleen, Christel G. Rutte, Brend Seinen, and Michiel Kompier (2001), "Autonomous Teamwork and Psychological Well-Being," European Journal of Work and Organizational Psychology , 10 (3), 291-301.

Van Yperen, Nico W. and Tom A.B. Snijders (2000), "A Multi-level Analysis of the Demands-Control Model: Is Stress at Work Determined by Factors at the Group Level or the Individual Level?" Journal of Occupational Health Psychology , 5 (1), 182-90.

Wageman, Ruth (1995), "Interdependence and Group Effectiveness," Administrative Science Quarterly , 40 (1), 145-80.

----- (1997), "Case Study: Critical Success Factors for Creating Superb Self-Managing Teams at Xerox," Compensation and Benefits Review , 29 (5), 31-41.

Yeatts, Dale E. and Cloyd C. Hyten (1998), High-Performing Self-Managed Teams: A Comparison of Theory to Practice . Thousand Oaks, CA: Sage Publications.

Zemke, Ron (1993), "Rethink the Rush to Team up," Training , 30 (11), 55-61.

~~~~~~~~

By Ad de Jong; Ko de Ruyter and Jos Lemmink

Ad de Jong is a postdoctoral researcher (e-mail: a.dejong@mw.unimaas.nl), Maastricht Academic Center for Research in Services

Ko de Ruyter is Professor of International Service Research and Director of the Maastricht Academic Center for Research in Services (e-mail: k.deruyter@mw.unimaas.nl)

Jos Lemmink is Professor of Marketing and Market Research and Department Chair (e-mail: j.lemmink@ms.unimaas.nl), Department of Marketing, Faculty of Economics and Business Administration, Maastricht University, The Netherlands.

Antecedents and Outcomes of Marketing Strategy Comprehensiveness 

Comprehensiveness has long been recognized as a key feature of marketing strategy decision making. However, few studies have examined its antecedents and the conditions under which it influences performance. This study attempts to contribute to a better understanding of marketing strategy by investigating project-level antecedents and outcomes of marketing strategy comprehensiveness (MSC). Drawing on contingency and institutional theories perspectives, the authors develop and test the effects of output and process rewards, task conflict, and project members' intra-and extra-industry relationships on MSC. They find that whereas process reward and extra-industry relationships are positively related to MSC, task conflict (when combined with conflict avoidance) hinders its development. Furthermore, the results indicate that MSC has a more positive effect on performance when implementation speed is higher. Finally, the authors discover that technology and market uncertainties differentially moderate the relationship between MSC and performance; the former has a positive effect, and the latter has a negative effect. The authors also discuss the theoretical and practical implications of their findings.

For several years, marketing scholars have dedicated considerable efforts to understanding the process of marketing strategy development and implementation (Bonoma 1984; Burke 1984; Glazer and Weiss 1993; Menon, Bharadwaj, and Howell 1996; Menon et al. 1999; Noble and Mokwa 1999). A key feature of this process is marketing strategy comprehensiveness (MSC), defined as the extent to which project members are extensive and exhaustive in the search for market information, the generation of many alternative courses of action, the examination of multiple explanations, and the use of specific criteria in making decisions in marketing strategy development and implementation. Comprehensiveness reflects the structure, rigor, and thoroughness of information search and analysis in marketing strategy decision making (Menon et al. 1999). It forces managers into a hypothesis-testing mode (Eisenhardt 1989, p. 558), thereby creating new insights that ensure a better understanding of marketing strategy and an increase in managerial confidence in decision making (Day 1994). Therefore, comprehensiveness is a key component of a quality marketing strategy (Menon, Bharadwaj, and Howell 1996). Despite its importance, the antecedents and conditions under which MSC influences performance have yet to be addressed in extant theoretical or empirical research. There is a good reason they should be: Increasingly, marketing managers go to great lengths and employ several techniques (e.g., consideration of obligatory alternatives, fishbowl reviews) (Menon et al. 1999) in an attempt to be comprehensive in making a marketing strategy.

The purpose of this study is to address this oversight. The first research gap concerns the variability in the extent to which firms develop comprehensive marketing strategies. Andrews and Smith (1996) study marketing strategy creativity at the individual level and find that the manager's risk-taking propensity, diversity of education, knowledge of the environment, intrinsic motivation, and interaction with others are key antecedents. Menon and colleagues (1999) find that innovative culture is positively related to MSC measured at the firm level. Although these studies acknowledge that most marketing strategies are formulated and implemented for specific products, they do not focus on the project level. Thus, knowledge of the internal and external determinants of MSC at the project level is limited.

A second research gap pertains to the relationship between MSC and performance. Several studies have examined the direct effects of strategy comprehensiveness on performance and the moderating role of environmental uncertainty at the firm level. However, empirical findings have been mixed on both fronts (e.g., Fredrickson 1984; Fredrickson and Mitchell 1984; Goll and Rasheed 1997; Priem, Rasheed, and Kotulic 1995). The mixed findings may stem from two causes. First, by using the firm as a unit of analysis, previous studies ignore the array of internal and external factors that may influence the effect of a specific strategic decision process on performance (Hough and White 2003). Second, previous studies examine environmental uncertainty as a unidimensional construct, thereby ignoring the different sources of uncertainty and their potential differential implications for the efficacy of strategy comprehensiveness.

Against this backdrop, in this study, we focus specifically on a project level in an attempt to contribute to the marketing literature in three ways. First, drawing on contingency theory, we offer explanations for some of the variance in MSC. A key argument in contingency theory is that a better understanding of the nature of organizational strategies is gained by examining its antecedents in the form of internal organizational and external environmental factors (Hofer 1975). Given that MSC involves complex information-processing tasks, we focus specifically on factors such as reward systems and task conflict, which affect the information-processing capacity of project members as internal determinants of MSC.

Second, we draw on institutional theory to explain that pressures for conformity may influence how internal factors such as task conflict affect MSC and how project members' external relationships act as means by which they acquire external market information in developing MSC. Empirical investigation of these issues is extremely important because most previous studies focus on which internal and environmental factors affect marketing strategies (e.g., Glazer and Weiss 1993) but pay little attention to the processes by which such factors influence managers in developing or adapting their strategies (Zeithaml, Varadarajan, and Zeithaml 1988, p. 40).

Third, we examine the conditions under which MSC influences performance, defined as the extent to which a product, which is the focus of a marketing strategy, has achieved planned sales, market share, and profit objectives. Contingency theory also informs this issue by arguing that firms that can adapt their strategies effectively to both internal and external factors are likely to perform better (Hofer 1975; Zeithaml, Varadarajan, and Zeithaml 1988). We argue that given the costs involved in achieving a high level of MSC and possible diminishing returns to its benefits, it is likely that a level of MSC that is either too low or too high may hurt performance, but a moderate level may be optimal. This implies that managerial ability to develop a level of MSC that is consistent with organizational resources may determine its success. We examine the nonlinear relationship between MSC and performance that is implied by this contingency argument.

Scholars have long recognized that strategy implementation plays a potentially important role in the linkage between a marketing strategy and performance (Bonoma 1984; Noble and Mokwa 1999). We contribute to this tradition by investigating the moderating effect of implementation speed. Finally, Song and Montoya-Weiss (2001, p. 61) argue that "uncertainty should be studied in relation to specific components of the environment in order to properly attribute its effects." Following this insight, we advance the literature by investigating two sources of environmental uncertainty (technology and market uncertainties) in an attempt to uncover their differential moderating effects between MSC and performance. Through this richer framework and empirical assessment, we attempt to respond to Menon and colleagues' (1999, p. 36) call for research that "examines more complex models of antecedents and outcomes of [marketing strategy making]." Figure 1 presents the conceptual model.

Internal Antecedents of MSC

Drawing on contingency theory (Hofer 1975) and extant marketing strategy research (Andrews and Smith 1996; Burke 1984; Menon et al. 1999), we identify internal conditions that influence effective information processing in strategy making as determinants of MSC. Market information for strategy making tends to contain tacit components that retard collective interpretation and information sharing. Even when market information is made explicit in a codified routine and communicated to project members, some of them may not understand it because they interpret and apply the knowledge in a different functional and experiential context. Consequently, market information collection, analysis, and interpretation in marketing strategy making for a specific product is a complex process that involves the understanding of multiple functions (Menon, Bharadwaj, and Howell 1996; Menon et al. 1999). For this reason, development of a comprehensive marketing strategy requires internal processes that focus project members' attention on and commitment to effective market information processing (Noble and Mokwa 1999). Two internal processes that are capable of engendering effective generation and sharing of market information in marketing strategy development are reward systems (Burke 1984; Jaworski and Kohli 1993; Ruekert and Walker 1987) and task conflict among project members (Menon, Bharadwaj, and Howell 1996). Noting the potential importance of the former, Menon and colleagues (1999, p. 35) call for further research to explore how output-and process-based reward systems influence marketing strategy development. We respond to their call in this study.

Output reward refers to a process of monitoring and compensating project members for achieving desired performance targets, such as meeting deadlines, budgets, and target market success (e.g., sales volume, market share). Output reward provides incentives and responsibilities for results, thereby ensuring that project members eschew politics and commit to the strategy-making process. Reduced politics occurs with an increased focus on the content and objectives of the market strategy, whereas commitment occurs with the collective efforts of project members directed toward diverse information collection and in-depth analysis of strategic options (Noble and Mokwa 1999). In addition, output reward shifts performance risk from the organization to the project members because environmental and company factors that may affect the outcome of the marketing strategy are beyond the project members' control (Oliver and Anderson 1994). A key means of attenuating the risks is to increase the quality of strategy with thorough information search and evaluation. By allowing autonomy over the means of achieving outcomes, output reward offers an incentive for project members to be comprehensive in attenuating performance risk (Atuahene-Gima and Li 2002).

Process reward is a means of monitoring and compensating project members for completing specified procedures and activities that are critical to achieving desired objectives in marketing strategy development. A key objective of process reward is to ensure thorough market information search and analyses, particularly about customers and competitors at specific stages of the strategy-making process. When rewards are tied to the completion of such process activities, project members believe that they will be rewarded for the quality of their strategies. Thus, process reward increases members' propensity to find and discuss a wider range of marketing strategy options (Menon, Bharadwaj, and Howell 1996, p. 308). A characteristic of MSC is the diversity and extensiveness of search for alternative courses of action, which requires project members to explore beyond the firm's boundary. Because such exploration increases the risk of errors in strategy making, there must be some level of protection for project members. We argue that a process reward establishes a norm of risk sharing between the firm and project members, thereby encouraging them to be expansive and rigorous in marketing strategy making (Atuahene-Gima and Li 2002; Oliver and Anderson 1994). Therefore:

H[sub1]: The greater the (a) output reward and (b) process reward, the greater is the MSC.

The task conflict construct refers to the debate and disagreements among group members about the content, goals, and processes of marketing strategy development (Jehn 1995). Task conflict is a key resource in group decision making because the vigorous debate and disagreements among project members encourage them to gather new data, to delve into issues more deeply, and to gain a more complete and expansive understanding of problems to develop alternative solutions (Jehn 1995; Ruekert and Walker 1987). Despite these benefits, task conflict can be misinterpreted as personal criticism or misconstrued as a challenge to the competence of the project members, particularly by people who are perceived as experts on the issues under discussion (Pelled, Eisenhardt, and Xin 1999). This may limit the expansive and barrier-breaking thinking that is required for comprehensiveness. Institutional theory argues that organizational culture and political processes tend to perpetuate conformity within a firm or group. This suggests that given the tacitness of knowledge and the perceived power of experts, pressures to conform to dominant views and interpretations may render task conflict ineffective in enhancing MSC. On balance, then, the effect of task conflict on MSC is unclear.

Consistent with contingency theory, we posit that the impact of task conflict on MSC may depend on the conflict resolution strategy adopted (Ruekert and Walker 1987). Although there are several conflict resolution methods, Song, Xie, and Dyer (2000) suggest that collaboration and avoidance are the ideal and worst types, respectively, and are the types most often adopted by marketing managers. In this study, we focus on collaboration and avoidance for the sake of parsimony. Collaborative conflict resolution refers to the extent to which project members confront conflicts by openly exploring areas of differences and commonality to find integrative solutions that are in the best interests of the strategy-making process. Collaborative conflict resolution is likely to enhance the positive effect of task conflict on MSC for two reasons. First, it reduces the uncertainties and misattribution associated with task conflict because it ensures that people understand the concerns and perspectives of others. This enables project members to concentrate on the content of discussions rather than on personality issues (Song, Xie, and Dyer 2000). Second, collaborative conflict resolution creates a sense of common dependence, thereby enhancing the willingness of project members to share information and to explore strategic options, which in turn enhances MSC.

Conflict avoidance refers to the extent to which project members avoid, ignore, or smooth over conflicts. Thus, conflict avoidance describes behaviors that minimize conflicts explicitly. Conflict avoidance reduces productive interactions among project members and therefore severely limits the timely collection and use of accurate and quality information (Song, Xie, and Dyer 2000). In addition, conflict avoidance inhibits open communication and exchange of quality information because project members focus less on the content of dissenting information and more on the intentions and motivations of the members who provide the information. Consequently, project members tend to pursue self-interested motives that limit the scope of exploration of market information and strategic options (Jehn 1995).

H[sub2]: Task conflict has a more positive effect on MSC when collaborative conflict resolution is higher than when it is lower.

H[sub3]: Task conflict has a more negative effect on MSC when conflict avoidance is higher than when it is lower.

External Antecedents of MSC

Contingency theory also recognizes the importance of external environmental determinants of firm strategy (Hofer 1975). Given their cognitive and resource limitations, decision makers tend to rely on external referents for information and insight into plausible strategic alternatives (Cyert and March 1963). In this respect, institutional theory implies that managers' external ties serve as conduits for information that shape managerial views about the environment and the strategic choices they make. The key idea of this theory is that firm strategies and practices are embedded in social relationships and may have a social meaning. Consequently, managers are affected by conformity and legitimacy pressures to adopt prevailing strategies (DiMaggio and Powell 1983; Meyer and Rowan 1977). In support of this theory, Geletkanycz and Hambrick (1997) find that managers' external relationships with other managers within and outside their industry influence their propensity to conform to prevailing organizational strategies. Given the increasing focus on MSC among marketing managers (Menon et al. 1999), this suggests that project members' ties with managers within and outside the firm's industry are antecedents of MSC.

Intraindustry relationships are project members' ties with managers in the same industry as the focal firm. Through such relationships, project members gain more comprehensive knowledge of industry strategic norms and recipes and more insight into the nature and context of the marketing strategies of other firms. This allows for greater diversity of perspectives that enhance members' search and analysis of strategic alternatives. Extraindustry relationships are project members' ties with managers outside the focal industry. Extraindustry relationships provide project members with an even broader range of information about strategies of firms outside the focal industry (Geletkanycz and Hambrick 1997). Such relationships increase the strategic options considered for selection in marketing strategy making. Thus:

H[sub4]: The greater the (a) intraindustry relationships and (b) extraindustry relationships of the project members, the greater is the MSC.

Performance Effect of MSC

Nonlinear effect. A key argument in the extant literature is that strategy comprehensiveness enhances performance because by generating diverse information about the market environment and by identifying the strengths and weaknesses of several strategic options, the firm is in a better position to implement a strategy more effectively (Eisenhardt 1989, p. 558; Menon, Bharadwaj, and Howell 1996, p. 308; Menon et al. 1999, p. 26). Although there may be good reasons for this argument, there are also significant costs associated with MSC. Indeed, some scholars have argued that given the cognitive limitations and bounded rationality of decision makers, strategy comprehensiveness is nearly impossible because of the high cost and time-consuming processes of information acquisition and analysis (Bahaee 1992). These arguments imply that too much or too little MSC may diminish performance, thus suggesting that it has a positive effect on performance only at a moderate level. From a contingency theory perspective, this suggests that managers who are able to determine the level of MSC that reflects their cognitive and other internal resources will derive higher performance from MSC. Thus:

H[sub5]: There is an inverted U-shaped relationship between MSC and performance, such that at extremely high and low levels its effect on performance is negative, but at moderate levels its effect on performance is positive.

Moderating effect of implementation speed. Extant research argues that a marketing strategy is more likely to result in better performance when it is implemented successfully (Bonoma 1984; Noble and Mokwa 1999). Although strategy formulation cannot be completely disassociated from its implementation, such a conceptual distinction enables researchers to better identify the discrete, albeit overlapping, aspects of strategy making that in combination may affect performance (Eisenhardt 1989). A key aspect of successful strategy implementation is implementation speed, defined as the pace of activities between the time project members formulate a marketing strategy and the time they fully deploy it in the marketplace. Implementation speed captures the acceleration of the decision-making activities from their conception to their implementation. Implementation speed may enhance performance by itself because it implies an ability to deal with potential hindrances to the efficacy of a marketing strategy and to ensure first-to-market benefits for a product (Noble and Mokwa 1999). In addition to its potential direct effect on performance, we posit that implementation speed enhances the positive effect of MSC on performance. The logic is that the formulation of a comprehensive marketing strategy is inherently a slow process because it takes a lot of time to consider many alternatives; to obtain input from many sources; and to engage in extensive, in-depth analysis. Thus, MSC is likely to achieve positive performance effects if project members can implement it speedily. Speedy implementation enables the firm to tap quickly into the window of opportunities uncovered by the process of strategy making. For example, Eisenhardt (1989) finds that decision making in the most successful companies is simultaneously fast and comprehensive. In contrast, slow implementation may exacerbate the costs associated with being comprehensive, thereby diminishing the impact of MSC on performance.

H[sub6]: The effect of MSC on performance is more positive when implementation speed is higher than when it is lower.

Moderating effect of technology and market sources of uncertainty. Technology uncertainty is the speed of change and instability of the technological environment. The conventional wisdom is that technology information is highly time sensitive; that is, it becomes obsolete quickly (Weiss and Heide 1993). Such information is believed to be of a dense variety, reflecting a high frequency of unexpected and novel changes, which thus makes it difficult for firms to respond with objective and formal procedures. Technology information is also perceived as highly equivocal, which means that it has multiple and ambiguous underlying meanings and causes that defy specific analysis and uniform interpretation (Daft and Macintosh 1981). In addition, technology uncertainty tends to disrupt the balance between project resource needs and available firm resources and skills. As Song and Montoya-Weiss (2001) find, technology uncertainty disrupts synergies among project members' resources and skills and synergies needed for effective strategy making. This suggests that with the existence of technology uncertainty, MSC is likely to diminish performance.

A counterargument is that whereas technological uncertainty is time sensitive, it is nevertheless amenable to effective comprehensive strategic processes because it leads managers to increase their information search efforts (Weiss and Heide 1993). For example, Bourgeois and Eisenhardt (1988) find that successful firms adapt to rapid technological changes by adopting a strategic decision process that involves comprehensive information search and thorough analysis of strategic alternatives. The key argument in their findings is that the perceived time sensitivity of technology information is low because the direction of technological change tends to be predictable. As Pavitt (1998) shows, technological change tends to progress along defined trajectories, such that firms can recognize and understand the directions of change. In this respect, Glazer and Weiss (1993, p. 510) also argue that "higher levels of interperiod change that are predictable may not be troublesome" in marketing strategy making. This suggests that MSC becomes more important for performance when technology is uncertain, because project members can identify critical decision variables to allow for a more expansive and effective analysis of strategic options. These equivocal arguments suggest the following competing hypotheses:

H[sub7a]: Marketing strategy comprehensiveness has a more positive effect on performance when technology uncertainty is higher than when it is lower.

H[sub7b]: Marketing strategy comprehensiveness has a more negative effect on performance when technology uncertainty is higher than when it is lower.

Market uncertainty refers to the speed of change in competitor actions and customer needs and preferences (Jaworski and Kohli 1993). Market uncertainty involves significant pace of change, heterogeneity, and unpredictability of customer needs and competitor actions, all of which tend to curtail deliberate and expansive information search efforts and to defy precise and comprehensive analysis (Glazer and Weiss 1993). For this reason, project members require rapid and flexible strategic processes to enhance performance in such an environment. Yet MSC is time consuming and less flexible, which suggests that it is of little value to performance when market uncertainty is high, because strategic decisions quickly become irrelevant (Bahaee 1992, p. 210). In addition, firms typically collect customer and competitor information by analyzing customers' choice criteria and attribute comparisons of the firm's products with competitors' products (Day and Wensley 1988). Such practices tend to result in institutionalized analyses and responses in strategy making in highly uncertain market environments that diminish performance. In support of these arguments, Glazer and Weiss (1993) find that market information is highly time sensitive and is not conducive for obtaining effective MSC outcomes.

Although the preceding arguments are persuasive, there is a contrary viewpoint to them. Market uncertainty prompts firms to reach out to customers (Li and Calantone 1998), which leads to an enhanced understanding of emerging customer needs and competitor actions (Jaworski and Kohli 1993). This means that market information may be less equivocal to project members and therefore is amenable to specific analysis and interpretation. As Daft and Macintosh (1981, p. 208) argue, when information is analyzable, use of an objective analytical process in strategy making enhances performance because correct responses can usually be identified. In light of these divergent arguments, we posit the following competing hypotheses:

H[sub8a]: Marketing strategy comprehensiveness has a more positive effect on performance when market uncertainty is higher than when it is lower.

H[sub8b]: Marketing strategy comprehensiveness has a more negative effect on performance when market uncertainty is higher than when it is lower.

Sample and Data Collection

We drew our sample from a mailing list of U.S. manufacturing firms that we obtained from Thomson Directory. We made telephones calls to identify project managers who met two selection criteria: ( 1) were involved in and ( 2) were knowledgeable about marketing strategy decision making for the most recent product introduced to market by the firm. We identified 393 company informants who met our selection criteria. We ensured that the informants were professionally interested, conscientious, and committed to providing accurate data by assuring them of confidentiality and by offering them a summary of the results. Subsequent to two follow-up reminders, we received 149 usable questionnaires, for a response rate of 38%. Of the sample, 70% of respondents worked in the high-technology industry: information technology, computers, and software (20%); electronics and electrical and scientific equipment (20%); pharmaceutical and biotechnology (12%); and automotive components (18%). The rest worked in low-technology industries: food (8%); forest, paper, and building products (13%); and other (e.g., footwear, clothing) (9%). Following the work of Menon and colleagues (1999), we pooled the data because the analysis of variance test showed that the constructs did not differ significantly (p > .10) among the industry groupings. The average project size in the sample was 5.66 people (standard deviation [s.d.] = 3.95). The average market duration for the product (defined as the number of months the product has been offered for sale in the market) was 17.5 months (s.d. = 15.09). T-test analysis showed no significant differences (p > .10) in the study variables between early and late respondents, which suggests that nonresponse bias is not a major concern.

Previous studies (e.g., Li and Calantone 1998; Menon, Bharadwaj, and Howell 1996; Menon et al. 1999) have found that project managers in senior positions, such as chief executive officer, vice president, and marketing manager, are reliable sources of information about marketing strategies. Of our project manager informants, 55% listed their job titles as marketing manager, 9% as vice president of marketing, 11% as product manager, 17% as chief executive officer, and 8% as engineering manager. The average work experience of informants in their firms was 9.95 years. Their degree of involvement and knowledge about the marketing strategy on a ten-point scale (see Conant, Mokwa, and Varadarajan 1990) was a high 8.89. These characteristics of the informants imply that they had the knowledge and confidence to respond to the issues under study.

Measures of Constructs and Validity

Table 1 presents the measures and their sources. We pretested the instrument in interviews with 35 part-time MBA students who had a minimum of three years of business experience. We obtained feedback that pertained mainly to ambiguities or difficulties in responding to the items and suggestions for adaptations to ensure the clarity and appropriateness of items. We revised the instrument accordingly. We defined marketing strategy development as involving the determination of decisions (e.g., product design, development, promotion, pricing, distribution) that require large resource commitments and long time horizons and are difficult to reverse in the short run.

We measured performance (α = .84) with three items by asking respondents to indicate the extent to which the product has achieved its sales, market share, and profit objectives since its launch. We also asked respondents to indicate on a single item the degree to which the overall performance of the product has met management expectations (1 = "well below expectations," and 10 = "well above expectations"). These two measures have a high correlation of .72 (p < .001), which suggests that there is convergent validity. We measured MSC (α = .91) with five items by asking informants to rate, for example, the extent and depth of the search for strategic alternatives. We measured output reward (α = .89) with four items by asking informants to indicate the degree to which rewards for project members were based, for example, entirely on performance outcomes. We measured process reward (α = .87) with four items by asking informants, for example, the degree to which rewards were based on the quality of strategic decisions.

We measured task conflict (α = .88) with five items that tapped the degree of disagreements among project members about ideas, goals, and processes adopted in the strategy-making process. We measured collaborative conflict resolution (α = .81) with four items that examined the degree to which the project members confronted and collaborated to resolve conflicts in the strategy-making process. Similarly, we measured conflict avoidance (α = .72) with four items that reflected the extent to which the project members refrained from confronting the conflicts. We captured intraindustry relationships (α = .88) and extraindustry relationships (α = .96) with four new items based on Geletkanycz and Hambrick's (1997) conceptual descriptions and project members' contacts with managers within and outside the industry, respectively.

We measured implementation speed (α = .78) with four new items that tapped the degree to which the strategy implementation was timely and faster than the planned schedule. We measured technology uncertainty (α = .90) with four items that pertained to the unpredictability of changes in technology and the rate of product introductions. The five items measuring market uncertainty (α = .75) reflected the speed of change of customer demand and competitor actions.

We controlled for several variables. We used firm size, measured by the number of employees, to control for greater complexity and economies of scale in large firms in strategy making. We used project size, measured by the number of people who have significant influence in marketing strategy decision making, to control for interaction dynamics that affect performance of groups. We coded industry as high technology ( 1) and low technology (0). We included product advantage (α = .73) (measured with three items that reflected the quality of the product, compared with competitors' products and previous products of the firm) and market duration of the product (measured by the number of months the product has been on the market) because each is a likely antecedent of performance.

We sought to control for common method bias by encouraging respondents to seek multiple responses to the questionnaire. A t-test analysis indicated no significant differences for all the study variables between the 20% of questionnaires completed by multiple informants and the 80% of questionnaires completed by single informants. We conducted a statistical check for common method variance with the Harman one-factor method (Podsakoff and Organ 1986). If common method bias is a serious problem, a single factor should emerge or one general factor should account for most of the variance. A principal components factor analysis of all measures yielded 13 factors with eigenvalues greater than 1.0, with total explained variance of 73%. Because several factors were uncovered and the first factor accounted for only 15% of the variance, common method bias may not be a serious problem (Menon et al. 1999, p. 31). Finally, we tested several interaction effects that could not be explained by common method bias because informants could not have guessed the complex relationships involved (Aiken and West 1991; Evans 1985).

In a confirmatory factor analysis, each measure loaded significantly on the expected constructs, which demonstrates convergent validity. Together, the factor loadings and model fit indexes (goodness-of-fit index = .89, root mean square error of approximation = .06, comparative fit index = .91, and nonnormed fit index = .93) presented in Table 1 suggest that the model fit is acceptable. Table 2 reveals that the diagonal elements representing the square roots of the average variance extracted (AVE) for each of the constructs are greater than the off-diagonal elements, which satisfies the criterion of discriminant validity (Fornell and Larcker 1981; Sakar, Echambadi, and Harrison 2001). Finally, the constructs' previously reported alpha and the composite reliabilities (CRs) presented in Table 2 indicate that each exceeded the accepted reliability threshold of .70. Table 2 presents the correlations and descriptive statistics of the constructs. We used the average score of measures of each construct for further analysis.

Model Specification and Estimation

( 1) MSC = α[sub0] + α[sub1] (OUTR) + α[sub2](PROP) + α[sub3](TASKC) + α[sub4](COLLA) + α5(AVOID) + α[sub6](INTRA) + α[sub7](EXTRA) + α[sub8](TASKC X COLLA) + α[sub9](TASKC X AVOID) + α[sub10](CON[sub1-4]) + ε[sub1], and

( 2) PDPERF = β[sub0] + β[sub1](MSC) + β[sub2](IMPSD) + β[sub3](MKTUN) + β[sub4](TEKUN) + β[sub5](MSCSQ) + β[sub6](MSC X IMSPD) + β[sub7](MSC X MKTUN) + β[sub8](MSC X TEKUN) + β[sub9](CON[sub1-12]) + ε[sub2],

where

MSCSQ = squared term for marketing strategy comprehensiveness, OUTR = output reward, PROR = process reward, TASKC = task conflict, COLLA = collaborative conflict resolution, AVOID = conflict avoidance, INTRA = intraindustry relationships, EXTRA = extraindustry relationships, PDPERF = performance, IMPSD = implementation speed, MKTUN = market uncertainty, TEKUN = technology uncertainty, and CON = control variables.

As Aiken and West (1991) recommend, we meancentered relevant variables before we created the interaction terms. The variance inflation factors in the regression models were all less than 2, which indicates that multicollinearity is not a serious problem. Table 3 contains the results.

Model 1 in Table 3 tests the effects of the control variables on MSC. Model 2 adds the main effects of the antecedent variables, which contribute 42% (ΔF = 12.82, p < .001) more than the variance explained by the control variables. The addition of the interaction terms in Model 3 added 3% (ΔF = 2.60, p < .01) to the explained variance obtained in Model 2. H[sub1a], which predicts that output reward is positively related to MSC, is not supported. Process reward is positively related to MSC, in support of H[sub1b] (α = .24, p < .01). H[sub2], which suggests that the relationship between task conflict and MSC is more positive when collaborative conflict resolution is higher than when it is lower, is not supported. Rather, the results show that collaborative conflict resolution is a positive predictor of MSC (α = .15, p < .05). H[sub3] is supported because the relationship between task conflict and MSC is more negative when conflict avoidance is higher than when it is lower (α = -.16, p < .01). [H[sub4a], which posits a positive link between intraindustry relationship and MSC, is not supported. The link between extraindustry relationship and MSC is positive (α = .36, p < .001), in support of H[sub4b].

Model 4 in Table 3 reports the main effects of the control variables on performance. Note that in this model, we also controlled for process and output rewards (Atuahene-Gima and Li 2002), intra-and extraindustry relationships (Geletkanycz and Hambrick 1997), and task conflict and conflict resolution methods (Jehn 1995) because previous studies suggest that they can influence performance. Model 5 adds the main effects of the antecedent variables, which contribute 5% (ΔF = 1.73, p < .10) more than the variance explained by the control variables. Model 6 adds the squared term for MSC and the interaction terms. These variables increased explained variance by 8% (ΔF = 2.92, p < .01) more than the explained variance we obtained in Model 5. The data do not support H[sub5], which predicts a nonlinear relationship between MSC and performance (β = .07, not significant). In support of H[sub6], the data in Model 6 show that MSC has a more positive effect on performance when implementation speed is higher than when it is lower (β = .24, p < .001). H[sub7a] is supported because the relationship between MSC and performance is positively moderated by technology uncertainty (β = .18, p < .05). H[sub8b] is supported because market uncertainty negatively moderates the effect of MSC on performance (β = -.23, p < .01). Regarding the control variables, product advantage (β = .38, p < .001) and task conflict (β = .16, p < .10) are positively related to performance.

This study examines the antecedents and conditions under which MSC affects performance at the project level. The results suggest that by rewarding process activities in marketing strategy development, managers provide significant incentives for project members to broaden the search for strategic alternatives and to deepen their analysis in marketing strategy development. This finding builds on Menon, Bharadwaj, and Howell's (1996) finding that formalization is important for developing quality marketing strategy. It also provides evidence that formalized reward systems enhance strategy comprehensiveness (Fredrickson 1986). A key finding is that the positive effect of task conflict on MSC is completely buffered by conflict avoidance, such that it becomes negative. It appears that open debate, diversity of opinions, and interpretations of market information can often generate adverse reactions from people whose views are criticized (Jehn 1995). The avoidance of such conflicts stifles the positive effect of task conflict on MSC. It is surprising that the collaborative conflict resolution method does not moderate the link between task conflict and MSC, as we predicted. These findings suggest that conflict avoidance and collaborative conflict resolution methods are related phenomena but are not opposite ends of the same spectrum.

Extraindustry relationships, but not intraindustry relationships, significantly enhance MSC. By having extraindustry relationships, project members gain exposure to diverse information and perspectives, which help in the discovery and analysis of diverse alternative strategic options. Contrary to our expectations, intraindustry relationships have no relationship with MSC. A possible reason for this finding is that such relationships tend to limit project members' capacity to envision multiple and different strategic alternatives in marketing strategy making. Together, the findings illuminate the enhanced insights that can accrue from combining both contingency and institutional theories as an explanatory platform for MSC.

The results on the contingent relationship between MSC and performance advance the literature on three fronts. First, we bring some clarity into the literature by showing that the relationship between MSC and performance is not concave. Indeed, MSC has no significant, direct effect on performance in this sample. Second, MSC has a positive effect on performance when it is combined with implementation speed. Implementation speed, by itself, is negatively related to performance. This finding could suggest that because it is a slow process, MSC and implementation speed are contradictory, such that extreme comprehensiveness may have a strong dilutive effect on the benefits of speedy implementation. It is also possible that a firm has a relatively slow strategy development process but speeds up its implementation process to market a product quickly, thereby enhancing performance (Eisenhardt 1989).

We explored these arguments in a post hoc analysis. We created a typology of four groups of firms using the median split method and compared them using analysis of variance. Firms with high MSC and implementation speed (49 firms) had significantly higher performance mean scores (3.97) than did firms with low MSC and high speed (3.41) (38 firms), which in turn had higher mean scores than firms with high MSC and low speed (2.98) (32 firms), which in turn did not differ from the 18 firms that had low MSC and low speed (2.56) (F = 7.79, p < .01). The results suggest that though MSC is important, it must be implemented speedily to achieve positive performance effects, which reinforces the importance of strategy implementation in extant marketing strategy research (Noble and Mokwa 1999).

Third, on the premise that technology and market uncertainties may have different implications for the effect of marketing strategy making on performance, we examined their differential moderating effects. Technology uncertainty positively moderates the effect of MSC on performance, in support of the argument that it provides a more conducive environment for MSC because the direction of change may be recognizable. That is, the time sensitivity of technology information may be low and therefore amenable to systematic analysis and comprehension (Pavitt 1998). The argument that technology information is unsuited for effective MSC is therefore not supported by our data. In contrast, MSC diminishes performance when market uncertainty is high. This lends support to the view that market information is highly time sensitive and that strategy making requires more real-time information flow than can be provided through a formal, comprehensive process (Glazer and Weiss 1993).

In summary, our study provides empirical support for the argument that components of the environment differ in their perceived time sensitivity and information-processing demands on project members in marketing strategy making. This extends the work of Glazer and Weiss (1993) and indicates the perils of a simple categorization of the environment into certain and uncertain. Our study suggests that if distinctions are not made among sources of uncertainty, insights into the complexity of the moderating effect of environmental uncertainty on the relationship between MSC and performance may be obscured.

Managerial Implications

The findings of our study suggest that to engender MSC, managers must reward project members for adherence to specific processes that ensure rigor and thoroughness in information collection and analysis. In addition, managers should encourage project members to cultivate relationships with people outside their own industries to gain both new insights and an expanded perspective in marketing strategy development. This could involve a provision of specific training and resources that enhance the external relationships of project members. Finally, managers must note that avoiding conflicts that result from task disagreements may hurt the level of MSC.

The results of this study also caution managers that an unquestionable positive view may be too simplistic, because its impact on performance is moderated by both internal and external factors. Specifically, the study suggests the need for managers to pay more attention to enhancing their firm's strategy implementation capability, because speedy implementation appears to alleviate the costs of the inherent slowness of MSC. The differential moderating effects of technology and market uncertainties suggest to managers that these sources of uncertainties create different information-processing demands in strategic decision making. Therefore, managers must understand the nature of the information-processing expertise required under each environmental condition in undertaking MSC to increase its performance effects. These findings underscore the need for managers to be more proactive in training project members to acquire the appropriate information-processing skills for marketing strategy making.

Limitations and Directions for Further Research

This study has some limitations. First, although we followed previous research (Li and Calantone 1998; Menon, Bharadwaj, and Howell 1996; Menon et al. 1999) in using a single, knowledgeable project manager who held a senior-level position as our informant, the results may be subject to single-informant bias. A reasonable argument can be made that such senior managers' knowledge about a marketing strategy may be of a summary, top-line nature and likely reflects a positive bias. Further research that uses middle managers as informants would help clarify whether the results reported herein are sensitive to key informants' level of seniority.

Second, because the focus of our study was MSC at the project level, it cannot speak directly to the contradictory findings reported by studies of MSC at the firm level. However, we believe that future studies that test the thrust of our theoretical model at the firm level can provide a means for resolving the previous discordant findings. Third, although objective performance measures would have been more desirable, they are usually unavailable at the project level. However, we note that subjective measures of performance continue to be useful in studies of marketing strategy development (e.g., Menon, Bharadwaj, and Howell 1996; Menon et al. 1999) and may provide the best option given differences in the nature of industries, time horizons, economic conditions, and goals of the sample firms. Finally, the generalizability of the findings is limited because our sample is not representative of U.S. firms.

In addition to alleviating the limitations of this study, there are other fertile avenues for further research in this domain. First, further research should examine other project-level antecedents because we explained only 46% of the variance in MSC. Second, we find that collaborative conflict resolution does not moderate the effect of task conflict on MSC. Xie, Song, and Stringfellow (1998) identify several other conflict resolution methods, such as competition, accommodation, compromise, and hierarchical methods. The extent to which such methods moderate the effect of task conflict on MSC should be examined in further research. Third, our study examined intra-and extraindustry relationships with a specific focus on managers. However, consultants, suppliers, and customers may be sources of input into marketing strategy development. Studying project members' relationships with these external constituents will enrich the understanding of how MSC emerges.

Fourth, future researchers should further explore the internal and external conditions under which MSC affects other performance outcomes. In particular, implementation factors such as strategic and role commitment (Noble and Mokwa 1999) and cross-functional integration (Menon et al. 1999) require investigation. Fifth, although several previous studies have reported differential moderating effects of sources of uncertainty in the relationship between marketing strategy and performance (e.g., Atuahene-Gima 1995; Glazer and Weiss 1993; Jaworski and Kohli 1993; Weiss and Heide 1993), few have offered a theoretical rationale for the observed differences. Given the importance of environmental uncertainty to the understanding of marketing phenomena, our distinction between the differential moderating roles of technology and market uncertainties is a valuable addition to the literature. The theoretical differences advanced herein hold promise for further research that examines how marketing strategy and other marketing capabilities are influenced differentially by these and other sources of uncertainty.

Finally, some scholars would argue that our description of MSC is an incomplete view of the reality of strategic decision making because it fails to recognize the cognitive limitations of managers and their resource limitations in searching for and interpreting information (Bahaee 1992; Cyert and March 1963). Such scholars would point to incremental decision making that involves experienced-based mental routines that produce answers automatically without apparent formal information search and evaluation. Further research should examine the determinants and outcomes of such decision-making processes in marketing strategy making.

The authors thank the three anonymous JM reviewers for their support and constructive comments on previous versions of this article. This article also benefited from comments and suggestions by seminar participants at the Center of Innovation Management and Organizational Change, Department of Management, City University of Hong Kong. The work described in this article was fully supported by Strategic Research Grant No. 7000750 from City University of Hong Kong.

Legend for Chart:

A - Measures and Sources
B - Description
C - Standardized Factor Loading
D - t-Value

 A

                          B                            C      D

Performance
CR = .80
AVE = .68
(Atuahene-Gima 1995)

    To what extent have the objectives for marketing
    the product been achieved with respect to

    • Sales                                      .85   12.34

    • Market share                               .78   10.53

    • Profit                                     .87   13.01

MSC
CR = 89
AVE = .79
(Menon et al. 1999)

    During the marketing strategy development
    process (product and associated marketing
    strategies), to what extent did the project
    members

    • Develop many alternative courses of
    action to achieve the intended objectives?        .88   12.80

    • Conduct multiple examinations of any
    suggested course of action the project members
    wanted to take?                                   .82   11.04

    • Thoroughly examine multiple explanations
    for the problems faced and for the
    opportunities available?                          .81   10.96

    • Search extensively for possible
    alternative courses of action to take advantage
    of the opportunities?                             .78    9.50

    • Consider many different criteria before
    deciding on which possible courses of
    action to take to achieve your intended
    objectives?                                       .79    9.88

Output reward
CR = .89
AVE = .68
(Atuahene-Gima and
Li 2002)

    To what extent do you agree with the following
    statements about the process of rewarding
    project members?

    • Rewards to project members were entirely
    related to achievement of performance
    objectives for project activities.                .84   10.44

    • Rewards for project members were entirely
    based on final outputs achieved.                  .81    9.56

    • The project members' rewards depended
    upon the market performance of the product.       .79    8.45

    • In rewarding the project members, primary
    weight was placed on objective criteria such as
    results achieved.                                 .76    7.86

Process reward
CR = .83
AVE = .59
(Atuahene-Gima and
Li 2002)

    To what extent do you agree with the following
    statements about the process of rewarding team
    members?

    • Rewards to project members were based on
    subjective criteria such as attributes of the
    product.                                          .76    7.89

    • Rewards to project members were based on
    effectiveness of implementation of the            .75    7.70
    strategy rather than results.

    • The rewards depended entirely on the
    quality of strategic decisions made rather than
    results.                                          .66    6.66

    • Project members were rewarded for
    completing major stages in the
    marketing strategy development process.           .65    6.06

Task conflict
CR = .89
AVE = .69
(Menon, Bharadwaj,
and Howell 1996;
Pelled, Eisenhardt,
and Xin 1999)

    To what extent did project members in the
    marketing strategy development disagree with
    each other about

    • Ideas concerning the best way to maximize
    the effectiveness of the marketing strategy.      .65    7.60

    • Ideas concerning the different goal
    priorities for the marketing strategy.            .75    8.86

    • The best way to ensure the success of the
    strategy.                                         .73    8.82

    • Which marketing objectives should be       .62    7.25
    considered more important.

    • Different approaches for developing and    .85   11.18
    implementing the strategy.

Collaborative conflict
behavior
CR = .82
AVE = .55
(Jehn 1995; Song,
Xie, and Dyer 2000)

    When conflicts arose among project members
    during the marketing strategy development
    process:

    • We tried to exchange complete and
    accurate information to resolve them.             .77   10.12

    • We played down our differences and
    emphasized our common interests.                  .47    5.69

    • We engaged in genuine collaborative
    effort to resolve them.                           .82   11.29

    • We discussed them, focusing on the
    common goals of the strategy.                     .85   12.15

Conflict avoidance
behavior
CR = .73
AVE = .41
(Jehn 1995; Song,
Xie, and Dyer 2000)

    When conflict arose among project members during
    the marketing strategy development process:

    • We refrained from arguments about
    the issues.                                       .42    4.63

    • We avoided the issues altogether.          .65    7.60

    • We tried to stay away from any
    disagreements.                                    .75    8.86

    • Our disagreements were swept under
    the carpet.                                       .72    8.47

Implementation speed
CR = .75
AVE = .44
(New scale)

    To what extent do you agree with the following
    statements about implementing the market
    strategy?

    • The implementation was faster than in
    other previous strategies.                        .75    8.73

    • The implementation was much faster than
    our planned schedule required.                    .69    7.70

    • The strategy was implemented in a
    shorter time than expected.                       .57    6.08

    • Speed of implementation of the strategy
    was a critical concern of the project members.    .66    7.00

Product advantage
CR = .74
AVE = .50
(Atuahene-Gima
1995)

    To what extent do you agree or disagree with
    the following statements about the product?

    • Quality of the product compared well
    with competitor products.                         .78    9.55

    • The product was of higher quality than
    competing products.                               .59    6.17

    • Quality of the product compares well
    with others we have developed in the past.        .83   10.21

Market uncertainty
CR = .72
AVE = .37
(Jaworski and Kohli
1993)

    Indicate your degree of agreement about how
    well these statements describe the market
    environment for the product.

    • Competitor activities in the market
    are quite uncertain.                              .40    4.80

    • The product-market has many new
    competitors.                                      .45    5.26

    • The product-market competitive
    conditions are highly unpredictable.              .52    5.88

    • Customers' preferences change quite
    rapidly.                                          .77    7.22

    • Customers' needs in our industry are
    changing quite rapidly.                           .85    7.98

Technological
uncertainty
CR = .88
AVE = .65
(Jaworski and Kohli
1993)

    Indicate your degree of agreement about how
    well these statements describe the
    technological environment for the product.

    • The technology in the market
    environment was changing rapidly.                 .78   10.55

    • Technological changes provide big
    opportunities in the industry.                    .63    7.87

    • A large number of new product ideas
    have been made possible through technological
    breakthroughs in the industry.                    .81   11.11

    • There are major technological
    developments in the industry.                    .94   14.53

Intraindustry
relationships
CR = .84
AVE = .62
(New scale)

    To what extent do you agree with the
    following statements about your project
    members during the market strategy
    development?

    • Project members communicated
    frequently with knowledgeable executives
    within our industry.                              .59    8.33

    • Project members had close interactions
    with knowledgeable people about conditions in
    our industry.                                     .64    8.55

    • Project members received a lot of
    information from other executives within
    our industry.                                     .84   12.46

    • Project members received advice about
    the project from knowledgeable people within
    our industry.                                     .74   10.37

Extraindustry
relationships
CR = .92
AVE = .71
(New scale)

    To what extent do you agree with the following
    statements about your project members during the
    marketing strategy development?

    • Project members put a lot of effort into
    communicating with knowledgeable people outside
    our industry.                                     .70    9.23

    • Project members maintained close contacts
    with knowledgeable people in firms outside our
    industry.                                         .84   11.18

    • Project members learned a lot from
    knowledgeable people in firms not operating in
    our industry.                                     .71    9.30

    • Project members received useful
    information from knowledgeable people outside
    our industry.                                     .81   10.89

Notes: We measured all items on a five-point scale (1 = "No
extent," and 5 = "To a great extent"), except for performance,
which we measured on a five-point scale where 1 = "Strongly
disagree," and 5 = "Strongly agree."

Legend for Chart:

A - Variables
B - 1
C - 2
D - 3
E - 4
F - 5
G - 6
H - 7
I - 8
J - 9
K - 10
L - 11
M - 12
N - 13
O - 14
P - 15
Q - 16
R - 17

     A
               B          C          D          E          F
               G          H          I          J          K
               L          M          N          O          P
                                                Q          R

1. Performance
            .82(a)

2. MSC
            .21(*)     .89

3. Process reward
            .05        .34(**)     .77

4. Output reward
            .18(*)     .18(*)      .24(**)    .82

5. Intraindustry relationships
            .03        .21(*)      .15       -.01         .84

6. Extraindustry relationships
            .11        .44(**)     .09        .04         .22(**)
            .78

7. Task conflict
            .11        .36(**)     .26(**)    .02         .11
            .28(**)    .69

8. Collaborative conflict resolution
            .16        .23(**)     .16(*)     .14         .04
            .04       -.00         .74

9. Conflict avoidance behavior
           -.11       -.20(*)     -.01       -.24(**)    -.04
            .01       -.01        -.18(*)     .64

10. Implementation speed
           -.10        .14         .18(*)     .45(**)     .05
            .03        .22(**)     .31(**)   -.08         .66

11. Technology uncertainty
            .01        .00         .20(*)     .07         .02
            .09        .22(**)    -.08        .06         .01
            .81

12. Market uncertainty
            .14        .05         .08        .14        -.03
            .06        .35(**)    -.11        .12         .16
            .45(**)    .61

13. Product advantage
            .43(**)    .24(**)     .22(**)    .34(**)     .13
            .11       -.07         .38(**)   -.18(*)      .31(**)
           -.02       -.03         .71

14. Market duration of product
           -.06       -.12        -.11       -.02         .00
           -.06       -.10         .13       -.12        -.04
           -.13       -.23(**)    -.04       N.A.

15. Project size
            .07        .05         .07       -.01        -.03
            .01        .12         .13(*)    -.12         .20(*)
            .06       -.02         .01        .20(*)      N.A.

16. Firm size
            .08        .04         .05        .02        -.02
            .08        .16         .06       -.02         .10
            .14        .13         .08       -.02         .01
                                              N.A.

17. Industry dummy
            .07       -.02        -.13        .15         .03
            .11       -.09         .11       -.04         .10
            .25(**)   -.00         .16(*)    -.03        -.10
                                              .06         N.A.

Mean
           3.45       2.92        2.34       2.81        2.88
           2.24       2.84        3.71       2.26        2.97
           3.32       2.85        3.68      17.54        5.66
                                           401.15         .65

S.D.
            .91        .88         .86       1.03        1.05
           1.06        .85         .75        .85         .86
           1.10        .79         .77      15.09        3.95
                                           223.64         .35

(*) p < .05.
(**) p < .01.
(a) Figures on the diagonal are square roots of AVE.
N.A. = not applicable.

Legend for Chart:

A - Variables
B - Hypotheses
C - MSC Model 1
D - MSC Model 2
E - MSC Model 3
F - Performance Model 4
G - Performance Model 5
H - Performance Model 6

     A

               B              C                   D

                              E                   F

                              G                   H

Controls and Direct Effects

Project size

               -            .04                 .01
                           (.46)               (.19)

                            .01                 .05
                           (.13)               (.59)

                            .10                 .08
                          (1.08)               (.87)

Firm size

               -           -.02                -.05
                          (-.32)              (-.70)

                           -.04                 .03
                          (-.54)              (-.33)

                            .03                 .05
                           (.35)               (.52)

Industry dummy

               -           -.04                -.08
                          (-.49)             (-1.06)

                           -.06                 .05
                          (-.85)               (.60)

                            .05                 .05
                           (.57)               (.59)

Product advantage

               -            -                   -

                            -                   .42
                                              (4.12)(***)

                            .44                 .38
                          (4.28)(***)         (3.67)(***)

Market duration of product

               -            -                   -

                            -                  -.09
                                             (-1.07)

                           -.09                -.10
                         (-1.03)             (-1.17)

Technology uncertainty

               -           -.06                -.11
                          (-.57)             (-1.44)

                           -.10                 -
                         (-1.20)

                           -.09                -.05
                          (-.89)              (-.47)

Market uncertainty

               -            .09                 .06
                           (.92)               (.68)

                            .06                 -
                           (.77)

                            .18                 .10
                          (1.74)(*)            (.93)

Output reward

               H[sub1a]     -                   .05
                                               (.64)

                            .04                 .01
                           (.47)               (.13)

                            .07                 .13
                           (.71)              (1.22)

Process reward

               H[sub1b]     -                   .26
                                              (3.53)(***)

                            .24                -.07
                          (3.24)(**)          (-.82)

                           -.07                -.14
                          (-.72)             (-1.39)

Task conflict

               -            -                   .19
                                              (2.35)(**)

                            .17                 .15
                          (2.08)(**)          (1.53)(†)

                            .13                 .16
                          (1.27)              (1.55)(†)

Collaborative conflict
resolution behavior

               -            -                   .16
                                              (2.22)(**)

                            .15                 .04
                          (2.05)(*)            (.39)

                            .10                 .14
                           (.99)              (1.39)

Conflict avoidance behavior

               -            -                  -.16
                                             (-2.16)(**)

                           -.18                -.02
                         (-2.43)(**)          (-.17)

                           -.01                -.02
                          (-.13)              (-.20)

Intraindustry relationships

               H[sub4a]     -                   .05
                                               (.72)

                            .06                -.06
                           (.86)              (-.66)

                           -.05                -.03
                          (-.56)              (-.29)

Extraindustry relationships

               H[sub4b]     -                   .39
                                              (5.21)(***)

                            .36                 .02
                          (4.82)(***)          (.19)

                            .01                 .02
                           (.14)               (.17)

Implementation speed

               -            -                   -

                            -                   -

                           -.21                -.20
                         (-1.99)(**)         (-1.94)(*)

MSC

               -            -                   -

                            -                   -

                            .01                 .00
                           (.05)               (.03)

Relevant Interaction Effects

Task conflict x collaborative
conflict resolution behavior

               H[sub2]      -                   -

                            .02                 -
                           (.24)

                            -                   -

Task conflict x conflict
avoidance behavior

               H[sub3]      -                   -

                           -.16                 -
                         (-2.26)(**)

                            -                   -

MSC²

               H[sub5]      -                   -

                            -                   -

                            -                   .07
                                               (.71)

MSC x implementation speed

               H[sub6]      -                   -

                            -                   -

                            -                   .24
                                              (2.77)(***)

MSC x technology uncertainty

               H[sub7a]     -                   -
               H[sub7b]

                            -                   -

                            -                   .18
                                              (1.81)(*)

MSC x market uncertainty

               H[sub8a]     -                   -
               H[sub8b]

                            -                   -

                            -                  -.23
                                             (-2.11)(**)

R²

               -            .01                 .43

                            .46                 .22

                            .27                 .34

Adjusted R²

               -            .00                 .38

                            .39                 .13

                            .16                 .21

F-value

               -            .20                7.70(***)

                           7.14(***)           2.58(**)

                           2.42(**)            2.66(***)

ΔR²

               -            -                   .42

                            .03                 -

                            .05                 .08

Partial F-value

               -            -                 12.82(***)

                           2.60(**)             -

                           1.73(†)      2.92(**)

N

               -            135                 135

                            135                 131

                            131                 131

(†) p < .10.
(*) p < .05.
(**) p < .01.
(***) p < .001.

Notes: We report standardized regression coefficients (t-values
are in parentheses). Reduced sample size for Models 4-6 is the
result of the deletion of outlier cases.

DIAGRAM: FIGURE 1; Antecedents and Outcomes of MSC

Aiken, Leona S. and Stephen G. West (1991), Multiple Regression: Testing and Interpreting Interactions. Newbury Park, CA: Sage Publications.

Andrews, Jonlee C. and Daniel C. Smith (1996), "In Search of the Marketing Imagination: Factors Affecting the Creativity of Marketing Programs for Mature Products," Journal of Marketing Research, 33 (May), 174-87.

Atuahene-Gima, Kwaku (1995), "An Exploratory Analysis of the Impact of Market Orientation on New Product Performance: A Contingency Approach," Journal of Product Innovation Management, 12 (September), 275-93.

----- and Haiyang Li (2002), "When Does Trust Matter? Antecedents and Contingent Effects of Supervisee Trust on Performance in Selling New Products in China and the United States," Journal of Marketing, 66 (July), 61-81.

Bahaee, Mahmood (1992), "Strategy-Comprehensiveness Fit and Performance," Australian Journal of Management, 17 (2), 195-216.

Bonoma, Thomas V. (1984), "Making Your Marketing Strategies Work," Harvard Business Review, 62 (March-April), 69-76.

Bourgeois, L.J. and Kathleen M. Eisenhardt (1988), "Strategic Decision Processes in High Velocity Environments: Four Cases in the Microcomputer Industry," Management Science, 34 (July), 816-35.

Burke, Marian C. (1984), "Strategic Choice and Marketing Managers: An Examination of Business-Level Marketing Objectives," Journal of Marketing Research, 21 (November), 345-59.

Conant, Jeffery S., Michael P. Mokwa, and P. Rajan Varadarajan (1990), "Strategic Types, Distinctive Competencies, and Organizational Performance," Strategic Management Journal, 11 (9), 365-83.

Cyert, Richard M. and James G. March (1963), A Behavioral Theory of the Firm. New York: Prentice Hall.

Daft, Richard L. and Norman B. Macintosh (1981), "A Tentative Exploration into the Amount and Equivocality of Information Processing in Organizational Work Units," Administrative Science Quarterly, 26 (2), 207-224.

Day, George S. (1994), "The Capabilities of Market-Driven Organizations," Journal of Marketing, 58 (October), 37-52.

------ and Robin Wensley (1988), "Assessing Advantage: A Framework for Diagnosing Competitive Superiority," Journal of Marketing, 52 (April), 1-20.

DiMaggio, Paul and Walter W. Powell (1983), "The Iron Cage Revisited: Institutional Isomorphism and Collective Rationality in Organizational Fields," American Sociological Review, 48 (April), 147-60.

Eisenhardt, Kathleen M. (1989), "Making Fast Strategic Decisions in High-Velocity Environments," Academy of Management Journal, 32 (3), 543-76.

Evans, Martin G. (1985), "A Monte Carlo Study of the Effects on Correlated Method Variance in Moderated Multiple Regression Analysis," Organizational Behavior and Human Decision Processes, 36 (3), 305-323.

Fornell, Claes and David F. Larcker (1981), "Evaluating Structural Equation Models with Unobservable Variables and Measurement Error," Journal of Marketing Research, 18 (February), 39-50.

Fredrickson, James W. (1984), "The Comprehensiveness of Strategic Decision Processes: Extension, Observations, Future Directions," Academy of Management Journal, 27 (3), 445-66.

----- (1986), "The Strategic Decision Process and Organizational Structure," Academy of Management Review, 11 (2), 280-97.

----- and Terence R. Mitchell (1984), "Strategic Decision Processes: Comprehensiveness and Performance in an Industry with an Unstable Environment," Academy of Management Journal, 27 (2), 399-423.

Geletkanycz, Marta A. and Donald C. Hambrick (1997), "The External Ties of Top Executives: Implications for Strategic Choice and Performance," Administrative Science Quarterly, 42 (4), 654-81.

Glazer, Rashi and Allen M. Weiss (1993), "Marketing in Turbulent Environments: Decision Processes and the Time-Sensitivity of Information," Journal of Marketing Research, 30 (November), 509-521.

Goll, Irene and Abdul M.A. Rasheed (1997), "Rational Decision-Making and Firm Performance: The Moderating Role of Environment," Strategic Management Journal, 18 (7), 583-91.

Hofer, Charles W. (1975), "Toward a Contingency Theory of Business Strategy," Academy of Management Journal, 18 (4), 784-810.

Hough, Jill R. and Margaret A. White (2003), "Environmental Dynamism and Strategic Decision-Making Rationality: An Examination at the Decision Level," Strategic Management Journal, 24 (5), 481-89.

Jaworski, Bernard J. and Ajay K. Kohli (1993), "Market Orientation: Antecedents and Consequences," Journal of Marketing, 57 (January), 53-70.

Jehn, Karen A. (1995), "A Multimethod Examination of the Benefits and Detriments of Intragroup Conflict," Administrative Science Quarterly, 40 (2), 256-82.

Li, Tiger and Roger Calantone (1998), "The Impact of Market Knowledge Competence on New Product Advantage: Conceptualization and Empirical Examination," Journal of Marketing, 62 (October), 13-29.

Menon, Anil, Sundar G. Bharadwaj, and Roy Howell (1996), "The Quality and Effectiveness of Marketing Strategy: Effects of Functional and Dysfunctional Conflict in Intraorganizational Relationships," Journal of the Academy of Marketing Science, 24 (4), 299-313.

------, ------, Phani Tej Adidam, and Steven W. Edison (1999), "Antecedents and Consequences of Marketing Strategy Making: A Model and a Test," Journal of Marketing, 63 (April), 18-40.

Meyer, John W. and Brian Rowan (1977), "Institutionalized Organizations: Formal Structure as Myth and Ceremony," American Journal of Sociology, 83 (2), 340-63.

Noble, Charles H. and Michael P. Mokwa (1999), "Implementing Marketing Strategies: Developing and Testing a Managerial Theory," Journal of Marketing, 63 (October), 57-73.

Oliver, Richard L. and Erin Anderson (1994), "An Empirical Test of the Consequences of Behavior-and Outcome-Based Sales Control Systems," Journal of Marketing, 58 (October), 53-67.

Pavitt, Keith (1998), "Technologies, Products, and Organization in the Innovating Firm: What Adam Smith Tells Us and Joseph Schumpeter Doesn't," Industrial and Corporate Change, 7 (3), 433-52.

Pelled, Lisa H., Kathleen M. Eisenhardt, and Katherine R. Xin (1999), "Exploring the Black Box: An Analysis of Work Group Diversity, Conflict, and Performance," Administrative Science Quarterly, 44 (1), 1-28.

Podsakoff, Philip M. and Dennis W. Organ (1986), "Self-Reports in Organizational Research: Problems and Prospects," Journal of Management, 12 (4), 531-44.

Priem, Richard L., Abdul M. Rasheed, and Andrew G. Kotulic (1995), "Rationality in Strategic Decision Processes, Environment Dynamism, and Performance," Journal of Management, 21 (5), 913-29.

Ruekert, Robert W. and Orville C. Walker Jr. (1987), "Marketing's Interaction with Other Functional Units: A Conceptual Framework and Empirical Evidence," Journal of Marketing, 51 (January), 1-19.

Sakar, M.B., Raj Echambadi, and Jeffrey S. Harrison (2001), "Alliance Entrepreneurship and Firm Performance," Strategic Management Journal, 22 (June-July), 701-711.

Song, X. Michael and Mitzi M. Montoya-Weiss (2001), "The Effect of Perceived Technological Uncertainty on Japanese New Product Development," Academy of Management Journal, 44 (1), 61-80.

-----, Jinhong Xie, and Barbara Dyer (2000), "Antecedents and Consequences of Marketing Managers' Conflict-Handling Behaviors," Journal of Marketing, 64 (January), 50-66.

Weiss, Allen M. and Jan B. Heide (1993), "The Nature of Information Search in High-Technology Markets," Journal of Marketing Research, 30 (May), 220-33.

Xie, Jinhong, X. Michael Song, and Anne Stringfellow (1998), "Interfunctional Conflict, Conflict Resolution Styles, and New Product Success: A Four-Culture Comparison," Management Science, 44 (December), 192-206.

Zeithaml, Valarie A., P. Rajan Varadarajan, and Carl P. Zeithaml (1988), "The Contingency Approach: Its Foundations and Relevance to Theory Building and Research in Marketing," European Journal of Marketing, 22 (7), 37-64.

~~~~~~~~

By Kwaku Atuahene-Gima and Janet Y. Murray

Kwaku Atuahene-Gima is Professor of Innovation Management and Marketing, Department of Management, City University of Hong Kong (e-mail: mgkwaku@cityu.edu.hk). Janet Y. Murray is Associate Professor of International Business, Boeing Institute of International Business, John Cook School of Business, Saint Louis University (e-mail: murrayjy@slu.edu).

Antecedents of Export Venture Performance: A Theoretical Model and Empirical Assessment 

Both the size and the rapid growth of global exporting have focused the attention of marketing researchers on the factors associated with firms' export performance. However, knowledge of this increasingly important domain of marketing activity remains limited. To address this knowledge gap, the authors draw on the strategy and marketing literature to develop an integrative theory of the antecedents of export venture performance. The interplay among available resources and capabilities, competitive strategy decisions, and competitive intensity determines export venture positional advantages and performance outcomes in the theoretical model. The authors empirically assess predicted relationships using survey data from 287 export ventures. Results broadly support the theoretical model, indicating that resources and capabilities affect export venture competitive strategy choices and the positional advantages achieved in the export market, which in turn affect export venture performance outcomes. In contrast to structure-conduct-performance theory predictions, the data indicate that the competitive intensity of the export marketplace does not have a direct effect on export venture positional advantages or performance. However, competitive intensity moderates the relationship between export venture competitive strategy choices and the positional advantages realized.

Worldwide exporting has grown to exceed five trillion dollars annually and accounts for more than 10% of global economic activity (e.g., International Monetary Fund 2001; World Bank 2001). Because of increasing globalization, exporting is also a means of foreign market entry and sales expansion for firms; thus, it is a significant area of research interest in marketing (e.g., Cavusgil and Kirpalani 1993; Samiee and Anckar 1998). Researchers have responded to managers' and policymakers' interests by focusing attention on the internal (e.g., international experience, standardization of marketing programs) and external (e.g., industry technology, export market characteristics) antecedents of firms' export performance (e.g., Aaby and Slater 1989; Cavusgil and Zou 1994; Szymanski, Bharadwaj, and Varadarajan 1993). However, despite increased attention, theoretical and empirical knowledge of exporting remains limited and offers few insights for managers who are responsible for export performance and policymakers who are concerned with export trade development (Czinkota 2000; Katsikeas, Leonidou, and Morgan 2000).

The literature highlights three particular problems that limit existing research. First, the majority of studies either are descriptive and largely atheoretic (Axinn 1994; Katsikeas, Leonidou, and Morgan 2000) or draw on a wide range of divergent theoretical perspectives (Aaby and Slater 1989; Zou and Stan 1998). The resulting lack of a comprehensive theory base for explaining firms' export performance makes it difficult to integrate findings from different studies into a coherent body of knowledge (e.g., Aulakh, Kotabe, and Teegen 2000; Zou and Stan 1998). Second, the export venture (i.e., the firm's efforts in a single product or product line exported to a specific foreign market) has been identified as the primary unit of analysis in understanding export performance (Ambler, Styles, and Xiucum 1999; Myers 1999). Despite this, most studies adopt a firm-level unit of analysis and aggregate firms' various product-market export ventures (Katsikeas, Leonidou, and Morgan 2000; Madsen 1987), which makes it difficult to identify and isolate specific antecedents of export performance because firm-level analyses fail to capture differences in the strategies executed by export ventures that face various marketplace requirements (Ambler, Styles, and Xiucum 1999; Cavusgil and Zou 1994). Third, export performance is multidimensional, incorporating both economic and strategic dimensions (Bello and Gilliland 1997; Zou, Taylor, and Osland 1998). However, most studies use individual performance measures, such as the firm's export ratio, which may not represent important economic and strategic aspects of export performance (e.g., Cavusgil and Zou 1994; Shoham 1998). In addition, the many unrelated performance indicators used in different studies also make integration of empirical findings problematic (Aaby and Slater 1989; Aulakh, Kotabe, and Teegen 2000; Diamantopoulos 1998).

In this article, we address these three problems. We develop and empirically assess a comprehensive theory of export venture performance. Our study makes three contributions to knowledge in this increasingly important domain of marketing activity. First, we integrate the structure-conduct-performance (SCP) paradigm and the resource-based view (RBV), two rival theories that often have been viewed as incongruent, into a cohesive theoretical model of the antecedents of export venture performance. We present empirical evidence of the interplay between the resources and capabilities available to export ventures, competitive strategy decisions, and the competitive intensity of the export market served in determining export venture positional advantages and performance outcomes that support key relationships in our theoretical model. Our study provides an important new theory base on which to build further export research and to integrate findings, and it offers new insights for managers and policymakers. Second, our study provides new evidence as to how competitive intensity affects export venture performance, which has important implications for theory development. In contrast to SCP predictions, our results indicate that competitive intensity does not directly influence export venture positional advantage and performance outcomes. Rather, we show that competitive intensity moderates the relationship between the export venture's intended competitive strategy and its realized positional advantage. Third, our study provides empirical support for previously untested marketing theory propositions regarding the effects of resources and capabilities on business performance (e.g., Bharadwaj, Varadarajan, and Fahy 1993; Day 1994; Hunt and Morgan 1995). We also extend RBV theory in marketing by distinguishing between firms' resource endowments and the capabilities with which they are developed and deployed as sources of positional advantage (Teece, Pisano, and Shuen 1997).

We begin by integrating insights from established theory in economics and resource-based strategy with emerging theories of dynamic capabilities to develop a comprehensive theoretical model of key antecedents of export venture performance. Next, we draw on qualitative fieldwork and literature-based insights to specify relevant constructs. After describing our research design, we validate our measures and estimate a structural model that represents key relationships predicted by our theory. We then present results of the analysis and explore their theoretical, managerial, and policymaking implications. Finally, we assess the limitations of our study and suggest areas for further research.

Antecedents of Export Venture Performance

Two broad theoretical approaches, the SCP paradigm and the RBV, dominate explanations of firm performance. Of the few theory-based exporting studies that have been conducted, most have examined the antecedents of export performance from an SCP viewpoint (e.g., Aaby and Slater 1989; Axinn 1994; Cavusgil and Zou 1994; Zou and Stan 1998). The SCP paradigm posits that firm performance is determined primarily by two fundamental sets of antecedents. First is the structural characteristics of the firm's markets that determine the competitive intensity (or rivalry) the firm faces. In the export venture context, competitive intensity concerns the degree to which rivals in the target export marketplace are able and willing to respond to the actions of the firm's export venture (e.g., Jaworski and Kohli 1993; Porter 1980). The second antecedent is the firm's ability to achieve and sustain positional advantages through the efficient and effective execution of planned competitive strategy (Porter 1980, 1985; Scherer and Ross 1990). In this context, positional advantage pertains to the relative superiority of the export venture's value offering to customers in the target export market and the cost of delivering this realized value (Day and Wensley 1988; Porter 1985). Export venture competitive strategies are planned patterns of resource and capability deployments that support choices about how the venture will compete for target customers and achieve its desired goals (e.g., Aulakh, Kotabe, and Teegen 2000; Bharadwaj, Varadarajan, and Fahy 1993).

In contrast, the RBV emphasizes resources as central to understanding firm performance (e.g., Amit and Shoemaker 1993; Peteraf 1993). In this domain, recent theoretical contributions regarding dynamic capabilities distinguish between capabilities and other types of resources available to the firm (e.g., Makadok 2001; Teece, Pisano, and Shuen 1997). In the export venture context, resources are the firm-controlled asset stocks that constitute the raw materials available to the firm's export venture business units (e.g., Black and Boal 1994; Peteraf 1993). Capabilities are the organizational processes by which available resources are developed, combined, and transformed into value offerings for the export market (e.g., Amit and Shoemaker 1993; Day 1994). The RBV characterizes firms as idiosyncratic bundles of resources and capabilities that are available for deployment by the firm's business units (e.g., Conner 1991; Hamel and Prahalad 1994). Heterogeneity in the resources and capabilities explains variations in firm performance (Makadok 2001; Teece, Pisano, and Shuen 1997). From this perspective, export venture managers deploy available firm-specific resources and capabilities that result in positional advantage in the export market (Barney 1991; Grant 1991). Firms sustain an advantage if rivals are unable to acquire and deploy a similar or substitute mix of resources and capabilities (Dierickx and Cool 1989; Mahoney and Pandian 1992).

The SCP and RBV approaches historically have been positioned as competing theories that offer incongruent explanations of firm performance (e.g., Porter 1991; Spanos and Lioukas 2001). However, as we suggest in Figure 1, a dynamic view of business performance as a process (March and Sutton 1997; Van de Ven 1992), with identifiable stages and linkages between them, enables the two different viewpoints to be synthesized into a robust theoretical model of the antecedents of export venture performance.

Resources and Capabilities: Insights from RBV Theory

Consistent with RBV and dynamic capabilities theory, our theoretical model indicates that both the resources and the capabilities available to the export venture have a direct effect on the venture's positional advantage in its target export market (e.g., Collis 1995; Day and Wensley 1988). For example, reputational assets may translate directly into image-related positional advantages, and relationship-building capabilities may directly create relationship-based advantages (e.g., Day 1994; Srivastava, Shervani, and Fahey 1998). Resources, as inputs to the complementary capabilities with which available resources are combined and transformed to create value offerings, also have an indirect effect on positional advantage (e.g., Oliver 1997; Teece, Pisano, and Shuen 1997). For example, export-market-knowledge resources can be leveraged with complementary product development capabilities to create superior value offerings for the export market (e.g., Calantone, Schmidt, and Song 1996).

Competitive Strategy: Insights from SCP Theory

In addition to the direct resource and capability-positional advantage linkages, our theoretical model draws on SCP theory to posit that these relationships are mediated by the competitive strategy that the export venture pursues (e.g., Hunt 2000; Spanos and Lioukas 2001). Competitive strategy mediates the relationship between an export venture's available resources and capabilities and its positional advantage by determining ( 1) how well available resources and capabilities are matched with market requirements (Collis 1995; Teece, Pisano, and Shuen 1997), ( 2) the appropriateness of planned resource and capability allocations (Castanias and Helfat 1991; Oliver 1997), and ( 3) the quality of strategy implementation (Day and Wensley 1988; Dickson 1992). Thus, our theoretical model posits that both the strategic choices about how the venture will compete for target customers and the combinations of available resources and capabilities to be deployed in the export market mediate linkages among available resources and capabilities and the positional advantages achieved by the export venture (Conner 1991; Grant 1991).

Competitive Intensity: Insights from SCP and RBV Theories

A fundamental premise in SCP theory is that the structural forces that determine competitive intensity in a market have a strong impact on firm performance (McGahan and Porter 1997; Scherer and Ross 1990). Thus, SCP theory posits that the level of competitive intensity is an essential determinant of market attractiveness (e.g., Porter 1980, 1985), whereas RBV theory treats competitive intensity as a less significant issue. Nonetheless, RBV theory posits that rivals' willingness and ability to imitate a firm's strategy or to use substitute resources and capabilities to deliver an equal value proposition determine the extent to which a firm's positional advantage may be "competed away" (Barney 1991; Conner 1991; Dierickx and Cool 1989). Our theoretical model posits three ways that competitive intensity in the target export market affects export venture performance. First, rivals' ability to take independent strategic actions and to respond to the export venture's competitive strategy moves moderates the venture's success in translating its planned competitive strategy into realized positional advantages (cf. Jaworski and Kohli 1993). This may occur because the export venture fails to anticipate correctly rivals' independent strategic actions in planning its competitive strategy and because of rivals' responses to the venture's own competitive strategy moves. Second, because the export venture's positional advantage is relative to the positions of its rivals, competitive intensity also has a direct, negative impact on a venture's positional advantage (Cavusgil and Zou 1994). Third, competitive intensity affects the likelihood of price competition, the cost of achieving realized positional advantages (Porter 1980, 1985), and distributor and customer choices (Day and Wensley 1988). Thus, competitive intensity also directly affects export venture performance.

Positional Advantage and Export Venture Performance

Positional advantages are direct antecedents of export venture performance because the relative superiority of a venture's value offering determines target customers' buying behavior (Anderson, Fornell, and Lehmann 1994; Piercy, Kaleka, and Katsikeas 1998) and the outcomes of this behavior for the export venture (Cavusgil and Zou 1994; Karnani 1984). Our theoretical model considers both economic and strategic dimensions of export venture performance (Bello and Gilliland 1997; Zou, Taylor, and Osland 1998). Economic elements include the achievement of economic goals (e.g., sales and market share) and the resources consumed in doing so (cf. Peng and York 2001). Strategic performance elements include the achievement of other goals, such as the establishment and maintenance of relationships with key channel members in the target export market (Cavusgil and Zou 1994). Although both economic and strategic elements are important dimensions of export venture performance, there may be trade-offs between them in the short run (Aaby and Slater 1989). Therefore, our theoretical model posits that an export venture's positional advantage affects the venture's economic and strategic performance (Katsikeas, Leonidou, and Morgan 2000).

Dynamic Considerations: Insights from RBV and Dynamic Capabilities Theories

For any theory of business performance to be useful, it must be dynamic (Porter 1991). Our theoretical model is explicitly dynamic. We view the relationships predicted among resources, capabilities, competitive strategy, positional advantage, and performance as stages in the export performance process (March and Sutton 1997; Ven de Ven 1992). In addition, we posit two particular dynamic relationships in our theoretical model. First, RBV theory indicates that some of the economic outcomes of positional advantages will be reinvested to acquire or to develop available resources and capabilities (Grant 1991; Hamel and Prahalad 1994). In the same vein, marketing theory indicates that strategic outcomes, such as relationships with customers and channel members, often become "market-based assets" that add to the firm's existing resource stock (Srivastava, Shervani, and Fahey 1998). Second, RBV and dynamic capabilities theories indicate that because of learning effects, many resources and most capabilities are enhanced by use (Grant 1996). For example, export market knowledge and relation-shipbuilding capabilities are likely enhanced as a result of the experiential learning that is associated with their use in planning, executing, and monitoring the outcomes of competitive strategy decisions (Day 1994; Morgan et al. 2003).

Theoretical Model Summary

By integrating RBV and SCP predictions in a dynamic model of export performance, the main premise of our theoretical model (Figure 1) is that export ventures can achieve positional advantages in foreign markets and, in turn, superior performance by deploying available resources and capabilities while pursuing appropriate export venture competitive strategies. Furthermore, we theorize that competitive intensity in the export market directly affects the export venture's positional advantages and performance outcomes and limits its ability to execute competitive strategy decisions. Our theoretical model indicates that an export venture's performance is sustained over time by reinvestment, the creation of market-based assets, and learning effects that build and enhance the resources and capabilities available to the export venture.

As an integrative general theory of export venture performance, we conceptualized our theoretical model at the same level as the SCP and RBV theories on which it draws. Assessing relationships at this level of analysis required us to treat the variables in our model as higher-order constructs (e.g., Matsuno and Mentzer 2000; Zou and Cavusgil 2002). This necessitates the identification of relevant dimensions of the constructs in our model (Bagozzi 1994; Heide and John 1992). In addition, because the absence of relevant secondary data sources requires primary data collection to assess our theoretical model, we needed to identify or develop valid and reliable measures of each of the dimensions of our theoretical constructs. The difficulties of longitudinal data collection in export ventures preclude a time-series assessment of predicted dynamic relationships (Katsikeas, Leonidou, and Morgan 2000). However, cross-sectional primary data enable us to assess key relationships in our theoretical model. To this end, to identify specific dimensions of each of the higher-order constructs in our model and to guide the development of appropriate measures, we synthesized insights from exploratory fieldwork interviews and existing literature.

The fieldwork involved 17 in-depth interviews, each of which lasted one and a half to two hours, with marketing managers, international business development managers, chief executive officers, and account development managers in different firms from a cross-section of industries, including textiles, carpets, and rugs (Standard Industrial Classification [SIC] code 22); finished apparel (SIC code 23); chemicals, plastics, paints, and cosmetics (SIC code 28); rubber and plastic products (SIC code 30); engines, machinery, and computing and office equipment (SIC code 35); and electronic and electrical components and appliances (SIC code 36). Overall, the interviewed managers were responsible for 29 export ventures. In addition to providing insights into the identification and measurement of dimensions that represent our theoretical constructs, the fieldwork interviews also provided important support for the face validity of our theoretical model.

Resources Available to the Export Venture

Although many different kinds of resources may be available, four emerged as particularly important in our fieldwork. First, experiential resources, such as market and process knowledge gained from the firm's overseas market operations experience, enable the venture's marketing programs to match the needs of channel members and customers (Daily, Certo, and Dalton 2000; Morgan et al. 2003). Second, scale resources, which pertain to the size and scope of the firm's operations, significantly affect cost structures and influence competitive strategy and performance (e.g., Cavusgil and Zou 1994; Cooper and Kleinschmidt 1985). Third, the working capital and financial liquidity requirements of export operations mean that access to financial resources is essential (Gomez-Mejia 1988; Tseng and Yu 1991). Fourth, physical resources, such as modern equipment and access to valuable supply sources that facilitate process efficiency and product effectiveness, can also be important sources of advantage in export markets (Cavusgil and Naor 1987; Leonidou, Katsikeas, and Piercy 1998).

Capabilities Available to the Export Venture

The literature and fieldwork interviews suggest three particularly important types of capabilities. First, informational capabilities, which pertain to the acquisition and dissemination of information about customers, competitors, channels, and the broader export market environment, help reduce uncertainty in export marketing (Katsikeas and Morgan 1994; Souchon and Diamantopoulos 1996). Second, relationship-building capabilities (with suppliers, customers, and other channel members) enable better understanding of and response to export market requirements (Bello, Urban, and Verhage 1991; Rosenbloom and Larsen 1992). Third, product development capabilities, which include existing product modification and new product development, affect the venture's effectiveness and efficiency in delivering superior value to the target market (Calantone, Schmidt, and Song 1996; Cooper and Kleinschmidt 1985).

Export Venture Competitive Strategy

We identified three key areas of planned resource and capability deployment that support a venture's strategic choices in competing for target customers. First, cost leadership, such as investments in new manufacturing technologies, enhances efficiency in the delivery of value offerings to customers (Aulakh, Kotabe, and Teegen 2000; Hill 1988; Sullivan and Bauerschmidt 1991). Second, marketing differentiation, such as investments in promotional and brand development activities, enables the delivery of a distinctive value offering to customers (Samiee and Roth 1992; Styles and Ambler 1994). Third, service differentiation, such as the implementation of customer-service programs that offer higher levels of customer support than do competitors' programs, enhance customer value (Cavusgil and Zou 1994; Roth and Morrison 1992).

Positional Advantage in the Export Market

We identified three types of positional advantage that have particular relevance to export venture performance: ( 1) cost advantage, which involves the resources consumed in producing and marketing the venture's value offering and affects price and perceived value in the export market (cf. Kotha and Nair 1995); ( 2) product advantage, which denotes quality, design, and other product attributes that differentiate the venture's value offering from those of competitors (Kim and Lim 1988; Song and Parry 1997); and ( 3) service advantage, which includes service-related components of the value offering, such as delivery speed and reliability and after-sales service quality (cf. Li and Dant 1999).

Export Venture Performance

Theory indicates that important aspects of economic performance are effectiveness (i.e., the extent to which desired goals are achieved), efficiency (i.e., the ratio of performance outcomes achieved to the resources consumed), and adaptability (i.e., the export venture's ability to respond to environmental changes) (Walker and Ruekert 1987). The literature, along with our fieldwork, suggests that strategic elements of export venture performance center primarily on two different stakeholders: distributors and end-user customers (Cavusgil and Zou 1994; Peng and York 2001). In particular, export ventures often monitor their performance with respect to desired customer attitudes and behaviors (e.g., customer satisfaction and retention) and those of channel intermediaries (e.g., distributor loyalty) (Katsikeas, Leonidou, and Morgan 2000).

Export Market Competitive Intensity

Our fieldwork supports previous research that has identified the willingness and ability of rivals to respond to competitive moves in the export market as an important antecedent of export venture performance (Cavusgil and Zou 1994).

Manufactured exports account for the bulk of total world export trade (World Bank 2001). Therefore, we empirically assessed our theoretical model in a field study of manufacturing firms that operate in the same SICs as firms in our previous fieldwork. We excluded service firms and firms engaged in primary industries because of their idiosyncratic international expansion patterns, regulatory requirements, and performance characteristics (Zou and Cavusgil 2002). We used a multi-industry sample to increase observed variance and to strengthen the generalizability of the findings (e.g., Bello and Gilliland 1997; Samiee and Anckar 1998). Most firms in the SICs in our sample export through foreign distributors because this provides relatively easy and low-cost export market access (Bello and Gilliland 1997; Peng and York 2001). To facilitate the development of valid measures and provide greater control over extraneous sources of variation, we therefore focused on only firms that exported through foreign distributors (cf. Dutta, Heide, and Bergen 1999).

Measures

We began by combining fieldwork and literature-based insights to specify the domain of each of the 17 construct dimensions we identified and to develop items that could serve as indicators of each construct. A preliminary survey instrument was developed and then evaluated by nine academic researchers in international marketing and competitive strategy who served as expert judges to assess face validity. Next, to evaluate individual item content, clarity of instructions, and response format, we pretested the revised survey in a series of face-to-face settings with 12 export venture managers. The survey was further refined by means of the feedback and was then pretested by mail. No particular problems were detected with the survey instrument. The final questionnaire used multi-item measures with seven-point scales to measure all constructs. Appendix A provides the complete set of items as well as their scale anchors and reliability coefficients.

Sample and Data Collection

We drew a random sample from the Dun & Bradstreet database of 1000 manufacturing firms that were involved in exporting and that employed between 50 and 1000 full-time personnel. We contacted each firm by telephone to identify the firms that had export venture activities through overseas distributors for at least five years, to identify an appropriate key informant for the study, and to prenotify the firm of the research project. After multiple telephone calls and successive snowballing in many cases, we identified 601 people who were responsible for specific export ventures, who met the key informant knowledgeability requirements, and who were willing to complete our survey. Of the 399 firms excluded in this process, 11 could not be contacted because of incorrect contact details, 68 traded directly with only a few export customers, 96 had been engaged in export venture activities for less than five years, 11 were local export intermediaries, 22 had discontinued exporting or employed less than 50 personnel, 8 had ceased operations entirely, and 183 reported a corporate policy of nonparticipation in external studies.

A survey packet was mailed to each of the 601 key informants. Respondents were asked to provide information for a specific export venture of the firm in which only one foreign distributor had been employed to sell the focal product in the venture market for at least five years. This enabled us to control for potential confounding factors associated with the use of multiple foreign distributors in a particular export venture market (Bello and Gilliland 1997) and to collect data on established export venture activities, which is essential in studying export venture performance (Cavusgil and Zou 1994).( n1) To ensure variation in export venture performance, we developed three versions of the survey. One version asked informants to respond with regard to one of their more successful export ventures; the other two focused respectively on averagely successful or less successful export ventures (cf. Weiss, Anderson, and MacInnis 1999).( n2) The initial mailing, a follow-up postcard, and two further waves of surveys produced 332 responses. Of these, 21 failed our informant competency tests (discussed subsequently), 15 had excessive missing data (missing responses on three or more items on any single scale; see Fitzgerald et al. 1997), and 9 were from export ventures that used multiple foreign distributors. These responses all were dropped from subsequent analysis, leaving a data set that comprised observations from 287 export ventures, for a response rate of 48%. Key demographic characteristics of the 287 export ventures in our data set are presented in Appendix B.

To assess potential nonresponse bias, we compared early and late respondents with respect to various firm characteristics, including number of full-time employees, years of exporting, annual sales volume, age of the venture, number of export markets, key informant self-reported competency evaluation indicators, and the construct measures (Armstrong and Overton 1977). We detected no significant differences between early and late respondents. In addition, using secondary information on employee numbers and annual sales volume, we also compared the respondent firms and a group of 87 randomly selected nonparticipant firms. We found no differences between respondents and nonrespondents. We concluded that nonresponse bias was not a significant problem in our data.

In addition to the presurvey telephone screening to identify appropriate informants, we also conducted a post hoc check for respondent competency. We collected data that tapped each respondent's knowledge of the activities, strategies, and performance of his or her export venture and those of its main export market competitors (cf. Jap 1999), involvement with the export venture's foreign distributor (cf. Heide and John 1992), responsibility for export venture strategy decisions (cf. Weiss, Anderson, and MacInnis 1999), and confidence in answering the survey questions (Cannon and Perreault 1999). We eliminated from further analysis the 21 respondents who reported a score of less than 4 on the seven-point scales for any one of these items. In the final data set (n = 287), the mean informant scores were greater than 6.0 on seven-point scales for all items except knowledge of the venture's main competitors, which had a mean score of 5.43. This indicates a high level of competency among our key informants.

We validated the data collected from our key informants in several ways. First, we attempted to gather data from a second informant in each respondent venture. In export ventures, usually only one manager is responsible for and knowledgeable about the full range of each venture's plans and activities. Even so, in 34 cases we were able to collect data from a knowledgeable second informant. Interrater reports were positively correlated, at levels ranging from .34 (p < .05) for export venture positional advantage to .74 (p < .01) for export venture competitive strategy. Second, we gathered data about the distributor and customer dimensions of export venture performance from 22 overseas distributors used by export ventures in our sample.( n3) Correlations between export venture manager and distributor responses for these export ventures were .60 (p < .01) and .40 (p < .10) for the distributor and end-user customer dimensions of venture performance, respectively. In the absence of secondary data sources to validate the economic dimension of export venture performance, we subsequently contacted the companies in our sample and requested their cooperation in providing us with objective financial performance data. We were able to collect primary objective economic performance data on sales volume, market share, relative profit margins, and revenue from new products for 31 of the export ventures in our sample. We correlated these objective data with the relevant indicators of economic performance that we used in our export venture performance measure (ECON1-4, Appendix A) at the level of .89 (p < .01), .89 (p < .01), .88 (p < .01), and .81 (p < .01), respectively. Collectively, the three sets of additional data provide strong support for the validity of our key informant data.( n4)

Measure Validation

We purified our measures using exploratory factor analysis and reliability analysis. We retained items with high item-to-total correlations, high loadings on the intended factors, and no substantial cross-loadings. We then subjected the set of items to confirmatory factor analysis (CFA) to verify the hypothesized factor structure. With the exception of competitive intensity, we considered each construct in our theoretical model as representing a higher-order factor, with the observed items originating from first-order factors that in turn arise from a second-order factor (cf. Heide and John 1992). Given the number of parameters to be estimated, sample-size constraints (Bentler and Chou 1987) led us to divide our measures into three subsets of the most theoretically related variables (e.g., Kohli and Jaworski 1994; Moorman and Miner 1997). In each measurement model, we used the elliptical reweighted least squares estimation procedure, which yields unbiased parameter estimates for both multivariate normal and nonnormal data (Sharma, Durvasula, and Dillon 1989; Zou and Cavusgil 2002).

Measurement Model 1 in Table 1 estimates resources available to the export venture as a second-order factor that comprises experiential, scale, financial, and physical resources, and it estimates capabilities available to the export venture as a second-order factor comprising informational, relationship-building, and product development capabilities. Although the chi-square statistic of Measurement Model 1 is significant (χ²(222) = 463.66, p < .001), as might be expected given the sensitivity of the test statistic to sample size (Bagozzi and Yi 1988), the other fit indexes (comparative fit index [CFI] = .97; nonnormed fit index [NNFI] = .96; root mean square error of approximation [RMSEA] = .062; and average off-diagonal standardized residual [AOSR] = .064) suggest good model fit. In Measurement Model 2, Table 1, we estimated export venture competitive strategy as a second-order factor comprising cost leadership, marketing differentiation, and service differentiation, and we estimated positional advantage in the export market as a second-order factor comprising cost, product, and service advantage. This model also represents a close fit to the data (χ²(163) = 264.12, p < .001; CFI = .96; NNFI = .96; RMSEA = .047; and AOSR = .048). Measurement Model 3, Table 1, estimates export venture performance as a second-order factor comprising economic, distributor, and end-user customer performance, and it estimates competitive intensity as a separate first-order construct. The results also suggest good fit for this model (χ²(131) = 248.66, p < .001; CFI = .97; NNFI = .97; RMSEA = .056; and AOSR = .041).

All three measurement models support our conceptualization of the resources and capabilities available to export ventures, competitive strategy, positional advantage, and performance as separate second-order constructs and competitive intensity as a separate first-order construct.( n5) As is shown in Table 1, all factors and items load significantly on designated constructs, and there is no evidence of any cross-loading. Factor and item loadings all exceed .52, and all t-values are greater than 8.55, which provides evidence of convergent validity among our measures (Fornell and Larcker 1981). We assessed discriminant validity among all of our measures by using two-factor CFA models that involved each possible pair of constructs; we freely estimated and then constrained to one the correlation between the two constructs. In all cases, the chi-square value of the unconstrained model was significantly less than that of the constrained model, providing evidence of discriminant validity between all of our constructs (Bagozzi, Yi, and Phillips 1991).( n6) The descriptive statistics and correlations in Table 2 provide a general picture of the constructs and measures. All measures exhibit satisfactory levels of reliability, and coefficient alphas range from .71 to .90 (see Appendix A). Overall, we conclude that our constructs exhibit good measurement properties.